Independent measured model card
Qwen/Qwen3.5-0.8B-Base
Hugging Face source ↗ · JSON for agents
853.0M unique parameters (dc7cdfe2ee41)
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 / Qwen3.5 same-base cohort full · 1,172 items | 38.48% accuracy · max | dc7cdfe2ee41codesota-qwen35-arc-challenge-v1 | 36.3s | Inspect run · Manifest |
| ARC-Easy / Qwen3.5 same-base cohort full · 2,376 items | 70.54% accuracy · max | dc7cdfe2ee41codesota-qwen35-arc-easy-v1 | 54.7s | Inspect run · Manifest |
How fine-tunes compare with this base
Open the complete same-base comparison →
ARC-Challenge / Qwen3.5 same-base cohort
ARC-Easy / Qwen3.5 same-base cohort
