Calibration
Claw-Eval → BFCL v4
BFCL v4 is estimated from Claw-Eval with a linear curve fitted on 5 models measured on both: y = 2.8304·x + -1.1404, R² = 0.60, cross-validated error 8.6 pp. It is used for 20 estimates.
| Estimated model | Claw-Eval | BFCL v4 | Source |
|---|---|---|---|
| Claude Opus 4.5 | 59.6% | 54.6% | estimated ± 8.6 pp, low confidence |
| Claude Opus 4.6 | 70.4% | 85.2% | estimated ± 8.6 pp, low confidence |
| Claude Sonnet 4.6 | 67.8% | 77.9% | estimated ± 8.6 pp, low confidence |
| DeepSeek V3.2 | 40.2% | 0.0% | estimated ± 8.6 pp, low confidence |
| dots3-note Preview | 73.4% | 93.7% | estimated ± 8.6 pp, low confidence |
| Gemini 3 Flash | 49.2% | 25.2% | estimated ± 8.6 pp, low confidence |
| GLM-5 | 57.7% | 49.3% | estimated ± 8.6 pp, low confidence |
| GLM-5-Turbo | 55.8% | 43.9% | estimated ± 8.6 pp, low confidence |
| GLM-5V-Turbo | 53.8% | 38.2% | estimated ± 8.6 pp, low confidence |
| K-EXAONE 2.0 | 77.7% | 100.0% | estimated ± 8.6 pp, low confidence |
| MiMo-V2.5 | 62.3% | 62.3% | estimated ± 8.6 pp, low confidence |
| MiMo-V2-Omni | 45.2% | 13.9% | estimated ± 8.6 pp, low confidence |
| MiMo-V2-Pro | 57.8% | 49.6% | estimated ± 8.6 pp, low confidence |
| Ornith-1.0-35B | 69.8% | 83.5% | estimated ± 8.6 pp, low confidence |
| Ornith-1.0-397B | 77.1% | 100.0% | estimated ± 8.6 pp, low confidence |
| Ornith-1.0-9B | 63.1% | 64.6% | estimated ± 8.6 pp, low confidence |
| Ornith-1.5-35B-A3B | 72.5% | 91.2% | estimated ± 8.6 pp, low confidence |
| Ornith-1.5-397B | 81.4% | 100.0% | estimated ± 8.6 pp, low confidence |
| Ornith-1.5-9B | 66.5% | 74.2% | estimated ± 8.6 pp, low confidence |
| Qwen3.5 397B | 56.8% | 46.7% | estimated ± 8.6 pp, low confidence |