21 October 2026 · GTX 1060 6 GB · Packet Foundry

Predator media: Qwen Image 2.1 Turbo Q2 · matched Q4/Q8/Q2 comparison

TL;DR

10 retained outputs · 10 accepted results · 10 recorded attempts · median Factory trace 198.9s Technical delivery and media decoding do not establish visual or musical quality. Timings include adapter overhead; cache behavior differs between runs. Failed and cancelled attempts remain in the evidence.

10 retained outputs · 10 accepted results · 10 recorded attempts · median Factory trace 198.9s

Technical delivery and media decoding do not establish visual or musical quality. Timings include adapter overhead; cache behavior differs between runs. Failed and cancelled attempts remain in the evidence.

← Complete campaign · Read the article · Download evidence

Qwen Turbo: matched Q4, Q8 and Q2

Ten new Q2 originals match twenty unchanged Q4/Q8 originals. Q2 changes loading policy, reserve, process and instrumentation as well as quantization. Actual logical CUDA checks and WDDM observations do not prove uninterrupted physical dedicated residency. Compare visible quality separately from timing.

WDDM three-second coverage passed for 7/10 whole-graph windows and 9/10 generator windows. Largest successful-sample gaps: 6.406083s and 6.074722s respectively. Actual query errors and failed coverage remain retained; sampled maxima do not bound unobserved gaps.

Matched workloadQ4 nativeQ8 nativeQ2 native
Initial request; actual process/cache state320.246s541.410s217.040s
Same first brief, next seed95.234s93.391s90.960s
Eight new briefs, retained-runtime median199.712s446.477s196.433s

Exact matched evidence and residency observations

Pair 1 · seed 42

Q4: initial request; inherited state · Q8: initial request; inherited state · Q2: initial request; inherited state

Q4_K_M · 320.246s native
Q4_K_M: A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Q8_0 · 541.410s native
Q8_0: A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Q2_K · 217.040s native
Q2_K: A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Exact matched prompt

A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.

Pair 2 · seed 43

Q4: 34.322s from previous native completion to start · Q8: 24.958s from previous native completion to start · Q2: 22.769s from previous native completion to start

Q4_K_M · 95.234s native
Q4_K_M: A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Q8_0 · 93.391s native
Q8_0: A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Q2_K · 90.960s native
Q2_K: A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Exact matched prompt

A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.

Pair 3 · seed 44

Q4: 22.977s from previous native completion to start · Q8: 29.801s from previous native completion to start · Q2: 16.310s from previous native completion to start

Q4_K_M · 194.865s native
Q4_K_M: A turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph.
Q8_0 · 620.242s native
Q8_0: A turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph.
Q2_K · 190.117s native
Q2_K: A turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph.
Exact matched prompt

A turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph.

Pair 4 · seed 45

Q4: 16.828s from previous native completion to start · Q8: 30.086s from previous native completion to start · Q2: 14.407s from previous native completion to start

Q4_K_M · 196.417s native
Q4_K_M: Product photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background.
Q8_0 · 532.767s native
Q8_0: Product photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background.
Q2_K · 192.086s native
Q2_K: Product photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background.
Exact matched prompt

Product photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background.

Pair 5 · seed 46

Q4: 16.283s from previous native completion to start · Q8: 29.459s from previous native completion to start · Q2: 23.027s from previous native completion to start

Q4_K_M · 199.640s native
Q4_K_M: A cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads "MORNING BREAD" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph.
Q8_0 · 444.960s native
Q8_0: A cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads "MORNING BREAD" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph.
Q2_K · 192.568s native
Q2_K: A cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads "MORNING BREAD" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph.
Exact matched prompt

A cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads "MORNING BREAD" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph.

Pair 6 · seed 47

Q4: 21.517s from previous native completion to start · Q8: 890.033s from previous native completion to start · Q2: 14.767s from previous native completion to start

Q4_K_M · 199.412s native
Q4_K_M: A bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text.
Q8_0 · 384.654s native
Q8_0: A bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text.
Q2_K · 196.334s native
Q2_K: A bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text.
Exact matched prompt

A bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text.

Pair 7 · seed 48

Q4: 13.667s from previous native completion to start · Q8: 27.120s from previous native completion to start · Q2: 23.127s from previous native completion to start

Q4_K_M · 199.784s native
Q4_K_M: An editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism.
Q8_0 · 447.995s native
Q8_0: An editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism.
Q2_K · 196.532s native
Q2_K: An editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism.
Exact matched prompt

An editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism.

Pair 8 · seed 49

Q4: 14.441s from previous native completion to start · Q8: 25.972s from previous native completion to start · Q2: 27.689s from previous native completion to start

Q4_K_M · 210.023s native
Q4_K_M: A charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering.
Q8_0 · 398.832s native
Q8_0: A charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering.
Q2_K · 200.972s native
Q2_K: A charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering.
Exact matched prompt

A charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering.

Pair 9 · seed 50

Q4: 13.101s from previous native completion to start · Q8: 38.627s from previous native completion to start · Q2: 29.696s from previous native completion to start

Q4_K_M · 208.566s native
Q4_K_M: A detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs.
Q8_0 · 496.444s native
Q8_0: A detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs.
Q2_K · 204.134s native
Q2_K: A detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs.
Exact matched prompt

A detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs.

Pair 10 · seed 51

Q4: 10.771s from previous native completion to start · Q8: 36.492s from previous native completion to start · Q2: 22.358s from previous native completion to start

Q4_K_M · 223.135s native
Q4_K_M: Create a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases.
Q8_0 · 421.721s native
Q8_0: Create a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases.
Q2_K · 217.850s native
Q2_K: Create a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases.
Exact matched prompt

Create a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases.

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 42

Seed 42 · 220.797 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 42

Download PNG · 0.30 MiB

  • A coherent, larger frontal seated fox on the desk, with a visible tail at the left.
  • Orange fox, cream backdrop and soft light match; requested ceramic identity is weak.
  • Dense fine fur and visible whiskers make it read more like an animal or furry toy than ceramic.
  • Face, ears and paws are recognizable; no obvious duplicate limbs.
  • No text requested or visible.
  • Requested ceramic material is not convincing; fur and whiskers dominate.
Exact prompt
A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 42
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 19,
    "complexity": "Very short / retained runtime pair",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 220.797228,
    "comfy_server_seconds": 217.04,
    "stages_seconds": {
      "node_4_observer_receive_interval": 2.4689999999827705,
      "node_2_observer_receive_interval": 3.719000000040978,
      "node_3_observer_receive_interval": 104.53099999995902,
      "node_12_observer_receive_interval": 0.01600000000325963,
      "node_11_observer_receive_interval": 0.031000000017229468,
      "node_1_observer_receive_interval": 0.7339999999967404,
      "node_5_observer_receive_interval": 0.09399999998277053,
      "node_10_observer_receive_interval": 0,
      "node_9_observer_receive_interval": 0.01600000000325963,
      "node_6_observer_receive_interval": 73.31200000003446,
      "node_7_observer_receive_interval": 31.57799999997951,
      "node_8_observer_receive_interval": 0.5
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [],
    "same_runtime_retained": null,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": null,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": false
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 108,
      "inflight_sample_count": 107,
      "minimum_ram_available_bytes": 4141621248,
      "minimum_commit_available_bytes": 13704994816,
      "minimum_system_drive_free_bytes": 18544037888,
      "maximum_sampled_device_gpu_used_mib": 4826
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": false,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": false,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-01.png",
      "mime_type": "image/png",
      "size_bytes": 310561,
      "sha256": "e978ec354acaa94c74eae1fdd18769fbfbd05fd974d6649ce3264b77ec3a3888",
      "source_sha256": "06e56e7ca63ffdc6b0bcb1f0c535ca622987e6abc8c0068040d2707f5eaf13c8",
      "width": 512,
      "height": 512,
      "pixel_sha256": "94c49b07bb2e075a707e35f8627a3561a2f9108920536fa2979acfcc73c7b37b",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 43

Seed 43 · 93.614 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 43

Download PNG · 0.27 MiB

  • Coherent frontal toy-like fox with a raised tail on the left.
  • Orange fox, pale desk, cream backdrop and studio light are present.
  • Fine fur grooves and soft-looking tail suggest a toy or animal texture rather than unmistakable ceramic.
  • Large dark eyes and small paws create juvenile/cartoon proportions; limbs remain separate.
  • No text requested or visible.
  • Ceramic identity is ambiguous; proportions are more juvenile and toy-like than the two matched originals.
Exact prompt
A small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 43
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 19,
    "complexity": "Very short / retained runtime pair",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA small orange ceramic fox on a pale wooden desk, soft morning light, plain cream background, realistic studio photograph."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 93.614167,
    "comfy_server_seconds": 90.96,
    "stages_seconds": {
      "node_9_observer_receive_interval": 0.03200000000651926,
      "node_6_observer_receive_interval": 61.46799999999348,
      "node_7_observer_receive_interval": 29.219000000040978,
      "node_8_observer_receive_interval": 0.14100000000325963
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "10",
      "11",
      "12",
      "2",
      "3",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": true,
    "context": {
      "native_inter_run_gap_seconds": 22.769,
      "only_seed_changed_from_predecessor": true,
      "text_changed_from_predecessor": false
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 47,
      "inflight_sample_count": 46,
      "minimum_ram_available_bytes": 6524329984,
      "minimum_commit_available_bytes": 13441687552,
      "minimum_system_drive_free_bytes": 18542247936,
      "maximum_sampled_device_gpu_used_mib": 4830
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": true,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-02.png",
      "mime_type": "image/png",
      "size_bytes": 278667,
      "sha256": "97e7660bd40ab579083276c59de0a8ec65ebc05e4008f92bd9a0cb7aa48957a0",
      "source_sha256": "001ff3b2bbc0bbfffb826c42240e0a1f92c74bacd0f83a16cd1d93a112be0357",
      "width": 512,
      "height": 512,
      "pixel_sha256": "cc4962b6079506dbf4a5dbea5774f0c07ac21776220124967a67bc3db0599d10",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 44

Seed 44 · 190.665 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 44

Download PNG · 0.48 MiB

  • Recognizable snow-covered mountains above a turquoise lake and dark forest.
  • Lake, snowy mountains and yellow boat near shore are present.
  • Mountains and foreground are grainy; mountain lighting and silhouette differ substantially from Q4/Q8.
  • Yellow boat at the lower left is small, soft and hull-like, without a clearly visible canoe interior.
  • No text requested or visible.
  • Canoe is indistinct and partly near the frame edge; stippled mountains and soft shoreline reduce photographic detail.
Exact prompt
A turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 44
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 19,
    "complexity": "Short landscape",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA turquoise alpine lake beneath snowy mountains at sunrise, a yellow canoe near the shore, crisp natural landscape photograph."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 190.664548,
    "comfy_server_seconds": 190.117,
    "stages_seconds": {
      "node_3_observer_receive_interval": 99.64100000000326,
      "node_10_observer_receive_interval": 0.01600000000325963,
      "node_9_observer_receive_interval": 0,
      "node_6_observer_receive_interval": 61.73399999999674,
      "node_7_observer_receive_interval": 28.67200000002049,
      "node_8_observer_receive_interval": 0.046999999962281436
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 16.31,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 94,
      "inflight_sample_count": 93,
      "minimum_ram_available_bytes": 6258405376,
      "minimum_commit_available_bytes": 13511499776,
      "minimum_system_drive_free_bytes": 18542092288,
      "maximum_sampled_device_gpu_used_mib": 4830
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-03.png",
      "mime_type": "image/png",
      "size_bytes": 500057,
      "sha256": "744b7233b57986a86a14e51e390b2f340475d6b70076b5a3992fa8f3989633ce",
      "source_sha256": "1ec6561ac84611b8de11bf56bef85bee0888bb4f1a388056fad4a8715371b2b0",
      "width": 512,
      "height": 512,
      "pixel_sha256": "a14b4a59e809ea19212fab34ad7a8f315a96386fe2f18b7a6f41b2baa533f94c",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 45

Seed 45 · 192.886 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 45

Download PNG · 0.26 MiB

  • Coherent black teapot product shot with a copper side handle.
  • Main requested materials, neutral background and soft lighting are present; two separate linen napkins are not confidently identifiable.
  • Matte black body and copper-colored handle are distinct; the dark base looks more like textured fabric/mat than stone.
  • Conventional side handle, lid and short spout are coherent; no obvious fusion.
  • No text requested or visible.
  • Dark slate is not convincing and appears textile-like; two distinct napkins are not clearly countable.
Exact prompt
Product photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 45
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 31,
    "complexity": "Product materials",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nProduct photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nProduct photograph of a matte black ceramic teapot with a warm copper handle, resting on dark slate beside two pale linen napkins. Three-quarter view, softbox lighting, crisp material detail, uncluttered background."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 192.886377,
    "comfy_server_seconds": 192.086,
    "stages_seconds": {
      "node_3_observer_receive_interval": 101.14000000001397,
      "node_10_observer_receive_interval": 0.01600000000325963,
      "node_9_observer_receive_interval": 0,
      "node_6_observer_receive_interval": 62.20299999997951,
      "node_7_observer_receive_interval": 28.65600000001723,
      "node_8_observer_receive_interval": 0.06299999996554106
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 14.407,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 95,
      "inflight_sample_count": 94,
      "minimum_ram_available_bytes": 6501715968,
      "minimum_commit_available_bytes": 13526708224,
      "minimum_system_drive_free_bytes": 18541101056,
      "maximum_sampled_device_gpu_used_mib": 4826
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-04.png",
      "mime_type": "image/png",
      "size_bytes": 267845,
      "sha256": "003b572810983e196e3a33dbc0ddeabd3209007ef220735b694772310bc7bcc9",
      "source_sha256": "9a4c8e00e94434f3aa9bf500f8481643852f31d304f513e7e465dd3e7f64c974",
      "width": 512,
      "height": 512,
      "pixel_sha256": "8e488b67bdecfcb31861b87ea7962bcbbcd3b4f64a7e08e7f741a6220a24bf8a",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 46

Seed 46 · 192.931 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 46

Download PNG · 0.41 MiB

  • Recognizable warmly lit bakery on a wet evening with a slightly angled view.
  • Requested storefront, bicycle, warm reflections and exact sign are present.
  • Warm glass and wet paving are visible, with softer low-contrast detail and grain in the facade.
  • Door/window geometry is generally coherent; bicycle frame and wheels are hard to separate in the dark right corner.
  • MORNING BREAD is clearly readable and correctly spelled.
  • Bicycle details merge into dark overlapping curves; facade/door detail is softer than the matched originals.
Exact prompt
A cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads "MORNING BREAD" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 46
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 41,
    "complexity": "Readable sign",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads \"MORNING BREAD\" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA cozy neighborhood bakery storefront on a rainy evening. The main sign clearly reads \"MORNING BREAD\" in cream block letters. Warm light glows through the windows, a bicycle leans beside the door, wet paving reflects the orange interior, realistic street photograph."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 192.930979,
    "comfy_server_seconds": 192.568,
    "stages_seconds": {
      "node_3_observer_receive_interval": 101.5,
      "node_10_observer_receive_interval": 0.015000000013969839,
      "node_9_observer_receive_interval": 0.01600000000325963,
      "node_6_observer_receive_interval": 62.375,
      "node_7_observer_receive_interval": 28.57799999997951,
      "node_8_observer_receive_interval": 0.061999999976251274
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 23.027,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 95,
      "inflight_sample_count": 94,
      "minimum_ram_available_bytes": 6338473984,
      "minimum_commit_available_bytes": 13460717568,
      "minimum_system_drive_free_bytes": 18527576064,
      "maximum_sampled_device_gpu_used_mib": 4826
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-05.png",
      "mime_type": "image/png",
      "size_bytes": 426742,
      "sha256": "b4c0eb114173be80b501ebd5f7d952c637a7cea88a2470c2f975e94b774b1038",
      "source_sha256": "6df09a9db58b9e8f024e1ade12ce471eae0d1532d70cf69a226d7b34598e2480",
      "width": 512,
      "height": 512,
      "pixel_sha256": "3af43f1640e218ec7223d4c46a7557eb91323e97dd81c7c3ca7c9e1acc7565bd",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 47

Seed 47 · 198.930 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 47

Download PNG · 0.64 MiB

  • Coherent sunlit greenhouse with benches, lush ferns, hanging pots and an open door.
  • Matches glasshouse, pots, plants and a narrow path; pronounced diagonal shadows are present. Hanging orchids are not clearly identifiable, and the brass object reads more as a bucket than a watering can.
  • Glass/iron and terracotta are distinguishable; foliage and window edges have conspicuous fine stippling.
  • Door and roof perspective remain recognizable; thin frame lines are sometimes wavy/doubled, path is mostly straight and a watering-can spout is not visible.
  • No text requested or visible.
  • Brass watering can is bucket-like without a visible spout; orchid blossoms are not resolved; window/plant edges are grainy.
Exact prompt
A bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 47
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 66,
    "complexity": "Botanical architecture",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA bright Victorian glass greenhouse filled with lush ferns, hanging orchids and terracotta pots. A narrow brick path curves from the bottom left toward an open green door at the far end. Morning sunlight makes geometric shadows through the iron window frames. Include a single brass watering can beside the path and misty trees beyond the glass. Architectural photograph, balanced natural colors, no people or text."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 198.930336,
    "comfy_server_seconds": 196.334,
    "stages_seconds": {
      "node_3_observer_receive_interval": 104.64100000000326,
      "node_10_observer_receive_interval": 0.031000000017229468,
      "node_9_observer_receive_interval": 0.01600000000325963,
      "node_6_observer_receive_interval": 62.51500000001397,
      "node_7_observer_receive_interval": 29.04700000002049,
      "node_8_observer_receive_interval": 0.09399999992456287
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 14.767,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 98,
      "inflight_sample_count": 97,
      "minimum_ram_available_bytes": 6357422080,
      "minimum_commit_available_bytes": 13414871040,
      "minimum_system_drive_free_bytes": 18539483136,
      "maximum_sampled_device_gpu_used_mib": 4826
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": false,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": false,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-06.png",
      "mime_type": "image/png",
      "size_bytes": 675033,
      "sha256": "40a896dce00a40bae66e1c0ba0c98acba66813cadb5696b49baf368ee8bae914",
      "source_sha256": "6579e3174710dc2d8fbc0f67001467f028769066646309ebf538abd9e3d03fdc",
      "width": 512,
      "height": 512,
      "pixel_sha256": "a7a39b3c95e1c6bb6163fa4a92d3ca5b159f5b27e1e1610299187818b38d01d5",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 48

Seed 48 · 198.943 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 48

Download PNG · 0.40 MiB

  • Recognizable adult artist holding a ceramic bowl in a pottery studio.
  • Blue apron, cream shirt, curly dark hair, bowl and softly blurred pottery shelves match; silver strands and precise rim grip are not distinct.
  • Face, linen and ceramic remain separable, but skin/hair and hands are softer than the two matched originals.
  • The bowl is supported partly underneath; fingers and hand edges blend together, and small facial/eye details are smeared.
  • No text requested or visible.
  • Finger separation is weak, especially under the bowl; facial/eye detail is soft; the requested silver strands and rim grip are not clearly shown.
Exact prompt
An editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 48
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 85,
    "complexity": "Portrait / hands",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nAn editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nAn editorial portrait of an adult female ceramic artist in her forties working in a sunlit pottery studio. She wears a blue linen apron over a cream shirt and calmly holds a small unfinished bowl in both hands, fingers naturally curled around the rim. Her face is clear and expressive, with a few silver strands in dark curly hair. Shelves of glazed pottery remain softly out of focus behind her. Frame from the waist upward, eye-level camera, gentle window light from the left, documentary realism."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 198.942577,
    "comfy_server_seconds": 196.532,
    "stages_seconds": {
      "node_3_observer_receive_interval": 105.09400000004098,
      "node_10_observer_receive_interval": 0.015000000013969839,
      "node_9_observer_receive_interval": 0.015999999945051968,
      "node_6_observer_receive_interval": 62.81200000003446,
      "node_7_observer_receive_interval": 28.5,
      "node_8_observer_receive_interval": 0.0779999999795109
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 23.127,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 98,
      "inflight_sample_count": 97,
      "minimum_ram_available_bytes": 6449410048,
      "minimum_commit_available_bytes": 13428981760,
      "minimum_system_drive_free_bytes": 18539282432,
      "maximum_sampled_device_gpu_used_mib": 4826
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-07.png",
      "mime_type": "image/png",
      "size_bytes": 419399,
      "sha256": "bc0c5eecd40aced92a3944789a8a53d98ad99e4b0c8d442bbc74369d244ee92a",
      "source_sha256": "791d4c49d4da135b6af93c0b8ef255fb6c52f59357da1a5e75b2ebc5bd40e103",
      "width": 512,
      "height": 512,
      "pixel_sha256": "5fd4fd0fe5f6aef34f48f9e650f5d98d1d12ac40a76948d2626780dbbc239190",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 49

Seed 49 · 201.829 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 49

Download PNG · 0.52 MiB

  • Readable brass robot in a tall old illustrated library, reaching toward a red book above a wooden stool.
  • Brass body, round glass-fronted head, red book, blue shoulder detail, stool and warm lamps are present. A distinct blue bird and ceiling constellations are not confidently resolved.
  • Brass, glass and wood are visually distinct in an etched/painterly style with strong line texture.
  • Arms are recognizable, but exact three-finger counts are unclear. Feet are not clearly visible on the stool; the lower legs merge into the region behind the seat.
  • No captions or readable lettering are visible.
  • Feet/standing support on the stool are ambiguous; exact three fingers per hand are not clearly shown.
  • Blue shoulder form is less clearly bird-like than the matched originals; constellations are not distinct.
Exact prompt
A charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 49
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 105,
    "complexity": "Stylized story scene",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA charming small brass robot with a round glass head stands on a wooden step stool inside an enormous old library, carefully returning a red book to a high shelf. The robot has two articulated arms and three visible fingers on each hand. A tiny blue mechanical bird perches on its shoulder. Bookshelves curve toward a domed ceiling painted with constellations; the far aisle is illuminated by warm amber lamps. Render as a detailed hand-painted storybook illustration with teal shadows, golden highlights and a clear focal point. Keep the robot, stool and book visibly separate, with no extra limbs, no captions and no written lettering."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 201.828697,
    "comfy_server_seconds": 200.972,
    "stages_seconds": {
      "node_3_observer_receive_interval": 105.82799999997951,
      "node_10_observer_receive_interval": 0,
      "node_9_observer_receive_interval": 0.015000000013969839,
      "node_6_observer_receive_interval": 66.25,
      "node_7_observer_receive_interval": 28.78200000000652,
      "node_8_observer_receive_interval": 0.0779999999795109
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 27.689,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 99,
      "inflight_sample_count": 98,
      "minimum_ram_available_bytes": 6506614784,
      "minimum_commit_available_bytes": 13419515904,
      "minimum_system_drive_free_bytes": 18554941440,
      "maximum_sampled_device_gpu_used_mib": 4824
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-08.png",
      "mime_type": "image/png",
      "size_bytes": 547451,
      "sha256": "45db3003f22fa429feff2b83ceee33b6d498d13add838a2233921922f9694616",
      "source_sha256": "e3065dea5f94eae6dc8cd9edbf08594f2595851b89489070a9ac0304d55d7edf",
      "width": 512,
      "height": 512,
      "pixel_sha256": "745fa8edc5e03c1920431a1a120e9607889e6f61d44ed8eaa42255590003ba1e",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 50

Seed 50 · 204.988 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 50

Download PNG · 0.45 MiB

  • Recognizable outdoor produce stall with three crates and several adults under a blue-and-white canopy.
  • Tomatoes, eggplants and lemons are separated into crates; vendor wears white and green outerwear is present. A large extra foreground figure obscures much of the right side and harbor; the requested paper-bag transfer is not distinct.
  • Produce and wood are recognizable, but canopy, faces and clothing have soft painterly/stippled detail rather than crisp photographic texture.
  • The exchanging hands merge into an indistinct cluster, with no cleanly separated bag. Three people appear, including an additional large green-jacketed back-facing figure.
  • No clearly readable signs or labels are visible.
  • Paper bag and hands are not cleanly separated.
  • Additional foreground person crowds the scene and obscures the harbor; exact two sailboats cannot be verified.
  • Fine face/canopy detail is soft and painterly.
Exact prompt
A detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 50
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 128,
    "complexity": "Crowded composition",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nA detailed travel photograph of a lively Mediterranean open-air produce market in early morning. In the foreground, three wooden crates hold red tomatoes, striped eggplants and yellow lemons, each type kept in its own crate. An adult vendor wearing a plain white shirt stands behind the stall, handing a small paper bag to an adult customer in a green jacket. A striped blue-and-white canopy casts cool shade over the stall. Beyond them, a narrow pedestrian street slopes toward a distant harbor with two white sailboats. Warm sunlight touches the upper walls but the foreground remains softly shaded. Compose the scene in layers with natural human proportions, believable hands, clear separation between people, baskets and produce, and rich but realistic color. Avoid artificial lens flares, watermarks and unreadable signs."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 204.987736,
    "comfy_server_seconds": 204.134,
    "stages_seconds": {
      "node_3_observer_receive_interval": 108.5,
      "node_10_observer_receive_interval": 0,
      "node_9_observer_receive_interval": 0.01600000006146729,
      "node_6_observer_receive_interval": 66.35899999993853,
      "node_7_observer_receive_interval": 29.125,
      "node_8_observer_receive_interval": 0.06299999996554106
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 29.696,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 101,
      "inflight_sample_count": 100,
      "minimum_ram_available_bytes": 6564458496,
      "minimum_commit_available_bytes": 13469601792,
      "minimum_system_drive_free_bytes": 18553696256,
      "maximum_sampled_device_gpu_used_mib": 4824
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": true,
    "wddm_generator_window_coverage": true,
    "wddm_observation_coverage": true,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-09.png",
      "mime_type": "image/png",
      "size_bytes": 471548,
      "sha256": "39c07e9e480d635b9e5a0023f87be29816db039fcce5ceda4299d8759bbdd500",
      "source_sha256": "28aee55010f8ea95464c8b2a262a0aa8f87523ee94ce2d5c32a8408f1f73694d",
      "width": 512,
      "height": 512,
      "pixel_sha256": "5d5f20c994fde16ebf5178872ed468f5c560150544c8a97aa6c24af220e6125d",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record

qwen21-turbo-q2-comparison · succeeded

Qwen Image 2.1 Turbo Q2_K · seed 51

Seed 51 · 219.911 s Factory trace

Generated output: Qwen Image 2.1 Turbo Q2_K · seed 51

Download PNG · 0.40 MiB

  • Recognizable quiet observatory at blue hour with one dome, red-coated figure, stairs and foreground pine/rocks.
  • Most major architectural and lighting elements are present, including the warm lamp and snow-covered steps. Sharp stars and a Milky Way band are not clearly visible; the small notebook-like object is not clearly closed.
  • Copper-colored dome, stone, snow and fabric remain distinguishable; thin edges and dark surfaces are strongly stippled.
  • One staircase and balcony are readable, but railings have wavy/overlapping fine lines and the figure/hand/notebook details are compressed.
  • No titles, labels or readable lettering are visible.
  • Requested stars and Milky Way are not resolved.
  • Balcony/stair railing detail is wavy and tangled; exact notebook state is unclear.
Exact prompt
Create a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases.
Parameters and prompt details
{
  "parameters": {
    "1 UnetLoaderGGUFAdvanced": {
      "unet_name": "qwen-image-2.1-turbo-Q2_K.gguf",
      "dequant_dtype": "float32",
      "patch_dtype": "float32",
      "patch_on_device": false
    },
    "2 CLIPLoader": {
      "clip_name": "qwen3vl_8b_int8_convrot.safetensors",
      "type": "qwen_image",
      "device": "cpu"
    },
    "3 TextEncodeQwenImage21": {
      "negative_prompt": "",
      "resolution": 512
    },
    "4 VAELoader": {
      "vae_name": "qwen_image_2.1_vae_bf16.safetensors"
    },
    "5 QwenImage21Cache": {
      "device": "off",
      "dtype": "default"
    },
    "6 PFQwenQ2WholeGeneratorSampler": {},
    "7 VAEDecode": {},
    "8 SaveImage": {},
    "9 RandomNoise": {
      "noise_seed": 51
    },
    "10 CFGGuider": {
      "cfg": 1.0
    },
    "11 KSamplerSelect": {
      "sampler_name": "euler"
    },
    "12 ManualSigmas": {
      "sigmas": "1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0"
    }
  },
  "prompt_details": {
    "word_count": 235,
    "complexity": "Elaborate constraints",
    "conditioning_texts": {
      "3.prompt": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nCreate a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases."
    },
    "actual_positive_text": "Create a single image from the supplied scene description.\n\nSource generation_brief:\nCreate a highly detailed cinematic architectural illustration of a remote mountain observatory at blue hour after a light snowfall. The building occupies the right two-thirds of the square image: a circular pale-stone base supports one weathered copper dome, whose open viewing slit points toward the upper left. A narrow exterior staircase with dark iron handrails climbs from a small lower entrance to a balcony; every step has a thin undisturbed layer of snow. One adult astronomer in a dark red winter coat stands on the balcony with a small white notebook held closed in one gloved hand, looking toward the sky. Place a warm amber lamp immediately above the entrance, creating a pool of light on the stone without illuminating the whole landscape. In the lower left foreground, a crooked pine tree and two exposed granite rocks frame a winding path that leads toward the building. Behind the observatory, show distant mountain ridges fading from slate blue into violet mist. The sky contains scattered sharp stars and a faint natural Milky Way band, with no oversized planets or fantasy moons. Balance warm interior light against cold ambient snow light, use physically plausible perspective, careful separation of the staircase rails and the figure, and visible differences between copper, stone, glass, fabric and snow. The final scene should feel quiet, inhabited and believable. Do not add titles, labels, logos, watermarks, extra people, extra domes or duplicate staircases."
  }
}
Timings, cache and resources
{
  "timing": {
    "factory_trace_seconds": 219.910558,
    "comfy_server_seconds": 217.85,
    "stages_seconds": {
      "node_3_observer_receive_interval": 116.32800000009593,
      "node_10_observer_receive_interval": 0,
      "node_9_observer_receive_interval": 0,
      "node_6_observer_receive_interval": 72.60899999993853,
      "node_7_observer_receive_interval": 28.82799999997951,
      "node_8_observer_receive_interval": 0.07900000002700835
    },
    "limits": "Native graph, trace and observer stage times have separate boundaries. Instrumentation adds overhead."
  },
  "cache": {
    "cached_nodes": [
      "1",
      "11",
      "12",
      "2",
      "4",
      "5"
    ],
    "same_runtime_retained": true,
    "immediate_warm_claim": false,
    "encoder_cached": false,
    "context": {
      "native_inter_run_gap_seconds": 22.358,
      "only_seed_changed_from_predecessor": false,
      "text_changed_from_predecessor": true
    }
  },
  "resources": {
    "sampled_summary": {
      "sample_count": 108,
      "inflight_sample_count": 107,
      "minimum_ram_available_bytes": 6545543168,
      "minimum_commit_available_bytes": 13439905792,
      "minimum_system_drive_free_bytes": 18553393152,
      "maximum_sampled_device_gpu_used_mib": 4846
    },
    "limits": "Device-wide observations include display/other workloads; exact per-process WDDM observations are separate."
  }
}
Encoder observations
{
  "encoder_cached": false,
  "token_truncation_verified": false
}
Verification and checksums
{
  "verification": {
    "full_decode": true,
    "checksum": true,
    "acknowledged": true,
    "trace_attachment": true,
    "native_experiment": true,
    "native_comparable_Q4_Q8_Q2_experiments": true,
    "independent_public_copy_decode": true,
    "visual_review": true,
    "logical_generator_placement": true,
    "wddm_evidence_validated": true,
    "wddm_whole_graph_coverage": false,
    "wddm_generator_window_coverage": false,
    "wddm_observation_coverage": false,
    "physical_dedicated_residency_proven": false
  },
  "media": [
    {
      "path": "media/qwen-turbo-q2-10.png",
      "mime_type": "image/png",
      "size_bytes": 414231,
      "sha256": "c8e09799d7d8e756e3221a2f7754525d504e0e0c355763120571b5e52d092a37",
      "source_sha256": "65c2fea2707a6ebc008ab1c6956867214f2f0a925fb51fe349e14a75738ae09c",
      "width": 512,
      "height": 512,
      "pixel_sha256": "1b62142ca270b4cad3505ea0f9692088965e9fd6eb75375625699ed2a9ca39cd",
      "metadata_handling": "PNG re-encoded losslessly without ancillary metadata; decoded pixels unchanged.",
      "public_copy_full_decode_passed": true,
      "source_content_preserved": "Decoded pixels byte-identical"
    }
  ]
}

Workflow snapshot

Download this run record