Independent model directory
Choose a model by its results.
Start with an evaluation. Compare measured performance, model size and execution cost, then inspect the evidence.
Selected evaluation
ARC-Easy / Qwen3.5 same-base cohort
Full pinned split · accuracy · higher is better
Best measured75.76%15 pinned systems
Protocol and scope
- Protocol
codesota-qwen35-arc-easy-v1- Scope
84fe51c6790b19376ad4a234c742b80e0f73e3eb0d6a7257d48e5315d6a7c078- Dataset
- allenai/ai2_arc · test · revision
210d026faf9955653af8916fad021475a3f00453 - Selection
- Full pinned split; configured limit 2376. Actual example counts appear beside every score.
Rows share the current evaluation scope. Different splits, prompts, scoring rules and older scopes are excluded. A subset score describes that subset; it is not a full-benchmark estimate.
| # | Model / pinned source | accuracy | Parameters | GPU-slot time | Evidence |
|---|---|---|---|---|---|
| 1 | Mapika/decider-0.8bCodeSOTA measurement · a0a01d6f8135 | 75.76%2376 examples | 752.4MCounted from pinned model | 1.5 min | Inspect run →Manifest |
| 2 | Qwen/Qwen3.5-0.8B-BaseCodeSOTA measurement · dc7cdfe2ee41 | 70.54%2376 examples | 853.0MCounted from pinned model | 54.7s | Inspect run →Manifest |
| 3 | unsloth/Qwen3.5-0.8B-BaseCodeSOTA measurement · 44bc982576bf | 70.50%2376 examples | 853.0MCounted from pinned model | 1.3 min | Inspect run →Manifest |
| 4 | NamanAgnih0tri/AlphaRoute-0.8B-v1.5CodeSOTA measurement · ef87d71f1105 | 69.61%2376 examples | 752.4MCounted from pinned model | 1.4 min | Inspect run →Manifest |
| 5 | McGill-NLP/AfriqueQwen3.5-0.8B-50LangsCodeSOTA measurement · f49099d79e51 | 67.68%2376 examples | 853.0MCounted from pinned model | 2.0 min | Inspect run →Manifest |
| 6 | mithulaartigala/reasoning-summarizer-qwen3.5-0.8b-previewCodeSOTA measurement · 558fc651bf01 | 66.92%2376 examples | 752.4MCounted from pinned model | 1.5 min | Inspect run →Manifest |
| 7 | Qwen/Qwen3.5-0.8BCodeSOTA measurement · 2fc06364715b | 65.87%2376 examples | 853.0MCounted from pinned model | 1.3 min | Inspect run →Manifest |
| 8 | Kewk/Heretical-Qwen3.5-0.8BCodeSOTA measurement · c16807de35d1 | 65.74%2376 examples | 853.0MCounted from pinned model | 1.4 min | Inspect run →Manifest |
| 9 | sh0ck0r/Qwen3.5-0.8B-hereticCodeSOTA measurement · fb35b60374fc | 65.19%2376 examples | 853.0MCounted from pinned model | 1.3 min | Inspect run →Manifest |
| 10 | ahmedeltawel/Qwen3.5-0.8B-hsb-dsbCodeSOTA measurement · 7e781a3e9704 | 63.59%2376 examples | 752.4MCounted from pinned model | 1.4 min | Inspect run →Manifest |
| 11 | cooperdk/BooruNL-0.8BCodeSOTA measurement · 14dbcf2e009f | 63.09%2376 examples | 853.0MCounted from pinned model | 1.7 min | Inspect run →Manifest |
| 12 | SeerRay-Lab/Xiaomi-OCR-0CodeSOTA measurement · ff990be318c4 | 56.52%2376 examples | 853.0MCounted from pinned model | 1.3 min | Inspect run →Manifest |
| 13 | mghafiri/qwen3.5-0.8B-decision-modelCodeSOTA measurement · 4a9939034a27 | 48.99%2376 examples | 752.4MCounted from pinned model | 1.5 min | Inspect run →Manifest |
| 14 | OmniJev/PlayJev-0.8BCodeSOTA measurement · a7348002b1e1 | 27.57%2376 examples | 853.0MCounted from pinned model | 1.6 min | Inspect run →Manifest |
| 15 | nineninesix/qwen3_5-full-attn-only-14CodeSOTA measurement · a26dff91d90e | 25.00%2376 examples | 511.2MCounted from pinned model | 46.1s | Inspect run →Manifest |
GPU-slot time is reserved evaluator wall time, including setup and CPU work; it is not kernel-active GPU time. Missing cost or parameter measurements stay unknown. Model-card claims and download counts do not become scores.