Calibration
SciCode → AA-SciCode
AA-SciCode is estimated from SciCode with a linear curve fitted on 20 models measured on both: y = 0.9244·x + 0.0319, R² = 0.90, cross-validated error 3.0 pp. It is used for 14 estimates.
| Estimated model | SciCode | AA-SciCode | Source |
|---|---|---|---|
| Sakana Fugu | 60.1% | 58.7% | estimated ± 3.0 pp, medium confidence |
| Sakana Fugu-Ultra | 58.7% | 57.4% | estimated ± 3.0 pp, high confidence |
| Kimi K2.5 | 48.7% | 48.2% | estimated ± 3.0 pp, high confidence |
| Beam 1.0 501B-A23B FP8 | 48.3% | 47.8% | estimated ± 3.0 pp, high confidence |
| Qwen 3.6 Max (preview) | 47.0% | 46.6% | estimated ± 3.0 pp, high confidence |
| Beam 1.0 501B-A23B NVFP4 | 46.9% | 46.5% | estimated ± 3.0 pp, high confidence |
| Sage-A1.5 | 44.3% | 44.1% | estimated ± 3.0 pp, high confidence |
| Ling 3.0 Flash FP8 | 40.4% | 40.5% | estimated ± 3.0 pp, high confidence |
| Spark-X2.6-4B | 36.8% | 37.2% | estimated ± 3.0 pp, high confidence |
| LFM2.5-24B-A2B | 32.7% | 33.4% | estimated ± 3.0 pp, high confidence |
| Nemotron 3 Nano Omni 30B A3B | 32.0% | 32.8% | estimated ± 3.0 pp, high confidence |
| Agents-A1-4B | 29.6% | 30.5% | estimated ± 3.0 pp, high confidence |
| LFM3-1.7B | 28.5% | 29.5% | estimated ± 3.0 pp, high confidence |
| Ling 2.6 Flash | 27.0% | 28.1% | estimated ± 3.0 pp, high confidence |
Fitted on
| Model | SciCode | AA-SciCode |
|---|---|---|
| A.X K2 | 41.0% | 41.0% |
| Gemini 3.5 Flash | 53.1% | 53.9% |
| Granite 4.2 30B | 38.8% | 37.8% |
| Granite 4.2 3B | 24.1% | 25.3% |
| Granite 4.2 8B | 36.1% | 31.5% |
| Grok 4.3 | 47.3% | 48.3% |
| Hy3 Preview | 41.2% | 48.6% |
| Inkling-Small | 48.7% | 49.7% |
| K-EXAONE 2.0 | 37.4% | 42.0% |
| Kimi K2.6 | 52.2% | 51.5% |
| Ling 3.0 Flash | 41.2% | 42.0% |
| Mercury 2.5 | 38.0% | 38.5% |
| MiniCPM5-2B | 26.3% | 26.3% |
| Muse Glimmer 30B | 43.6% | 44.9% |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 31.4% | 32.1% |
| Nemotron 3 Ultra | 44.6% | 40.3% |
| Qwen3.7 Max | 53.5% | 49.5% |
| Qwen3.7 Plus | 51.3% | 46.1% |
| Beam 1.0 501B-A23B | 49.2% | 48.5% |
| Step 5 Preview | 58.9% | 58.9% |