Archive · vps-81 historical telemetry · models/qwen2_5_coder_3b.html. Originally rendered 2026-04-16. Re-hosted from MyServers on 2026-05-06. Methodology and harness conventions may differ from what we use today; see /methodology.html for current standards. ← back to all benchmarks
Model Page

Qwen2.5 Coder 3B

TL;DR

This page indexes historical reports for Qwen2.5 Coder 3B. Labels such as “usable” describe the original harness checks, not independently verified correctness. Follow the linked reports for the test conditions.

qwen2.5-coder:3b

Back to overview Family qwen2.5-coder Size 3b Max context 32768
Details
Recorded runtime labels

ollama, ollama-local-mac, ollama-local-mac-throughput-sweep

Report count

20 linked report pages

Findings captured

11 summary lines from local JSON artifacts

Findings
  • Latest hello-check finished in 5.8s on ollama-local-mac.
  • Latest small-eval run averaged 67.2s per question, 3/5 strict passes, and 4.56 tok/s throughput.
  • Latest 10Q suite on ollama-local-mac averaged 31.1s for primary answers with 10/10 usable primary responses.
  • Latest 20Q suite on ollama-local-mac averaged 13.3s for primary answers with 20/20 usable primary responses.
  • Parallel lane 'Qwen2.5 Coder 0.5B plus Qwen2.5 Coder 3B' finished in 18.1s with 1.351x speedup and 26.0% wall savings on ollama-local-mac.
  • Parallel lane 'Qwen2.5 Coder 3B same-model 2-up' finished in 305.8s with 1.490x speedup and 32.9% wall savings on ollama-local-mac.
  • Parallel lane 'Qwen2.5 Coder 3B same-model 3-up' finished in 293.5s with 2.634x speedup and 62.0% wall savings on ollama-local-mac-throughput-sweep.
  • Parallel lane 'Qwen2.5 Coder 3B same-model 4-up' finished in 225.3s with 2.516x speedup and 60.2% wall savings on ollama-local-mac-throughput-sweep.
Benchmarks