Independent measured model card
Qwen/Qwen2.5-0.5B-Instruct
Hugging Face source ↗ · JSON for agents
494.0M unique parameters (7ae557604adf)
Every result links to its execution evidence. Compare scores only within matching protocols; no overall best-model claim. GPU-slot time includes evaluator setup and CPU work.
All current benchmark results
| Benchmark / coverage | Result | Version / protocol | GPU-slot time | Evidence |
|---|---|---|---|---|
| ARC-Challenge subset · 32 items | 40.62% accuracy · max | 7ae557604adfcodesota-sub1b-choice-v1 | 4.4s | Inspect run · Manifest |
| ARC-Easy subset · 32 items | 59.38% accuracy · max | 7ae557604adfcodesota-sub1b-choice-v1 | 4.5s | Inspect run · Manifest |
| ARC-Easy (full test split, raw accuracy) full · 2,376 items | 65.57% accuracy · max | 7ae557604adfcodesota-sub1b-arc-easy-standard-v2 | 33.4s | Inspect run · Manifest |
| BoolQ subset · 32 items | 71.88% accuracy · max | 7ae557604adfcodesota-sub1b-choice-v1 | 7.9s | Inspect run · Manifest |
| HellaSwag subset · 32 items | 40.62% accuracy · max | 7ae557604adfcodesota-sub1b-choice-v1 | 4.6s | Inspect run · Manifest |
| OpenBookQA subset · 32 items | 34.38% accuracy · max | 7ae557604adfcodesota-sub1b-choice-v1 | 4.2s | Inspect run · Manifest |