Calibration
CharXiv → Video-MME (with subtitle)
Video-MME (with subtitle) is estimated from CharXiv with a linear curve fitted on 5 models measured on both: y = 0.2024·x + 0.7121, R² = 0.88, cross-validated error 0.9 pp. It is used for 34 estimates.
| Estimated model | CharXiv | Video-MME (with subtitle) | Source |
|---|---|---|---|
| Claude Mythos 5 | 93.5% | 90.1% | estimated ± 0.9 pp, medium confidence |
| Claude Opus 4.5 | 68.5% | 85.1% | estimated ± 0.9 pp, low confidence |
| Claude Opus 4.7 (Adaptive) | 91.0% | 89.6% | estimated ± 0.9 pp, medium confidence |
| Claude Opus 4.8 | 89.9% | 89.4% | estimated ± 0.9 pp, medium confidence |
| Claude Sonnet 4.6 | 77.4% | 86.9% | estimated ± 0.9 pp, low confidence |
| Claude Sonnet 5 | 88.3% | 89.1% | estimated ± 0.9 pp, medium confidence |
| Command A+ | 52.7% | 81.9% | estimated ± 0.9 pp, low confidence |
| Gemini 3.1 Flash-Lite | 73.2% | 86.0% | estimated ± 0.9 pp, low confidence |
| Gemini 3.1 Pro | 80.2% | 87.4% | estimated ± 0.9 pp, medium confidence |
| Gemini 3.5 Flash | 84.2% | 88.3% | estimated ± 0.9 pp, medium confidence |
| Gemini 3.7 Flash | 88.7% | 89.2% | estimated ± 0.9 pp, medium confidence |
| Gemini 3 Pro | 81.4% | 87.7% | estimated ± 0.9 pp, medium confidence |
| GLM-5.3-Flash | 89.4% | 89.3% | estimated ± 0.9 pp, medium confidence |
| GPT-5.2 | 82.1% | 87.8% | estimated ± 0.9 pp, medium confidence |
| GPT-5.4 | 82.8% | 88.0% | estimated ± 0.9 pp, medium confidence |
| Grok 4.20 | 60.9% | 83.5% | estimated ± 0.9 pp, low confidence |
| Inkling | 82.0% | 87.8% | estimated ± 0.9 pp, medium confidence |
| Inkling-Small | 81.3% | 87.7% | estimated ± 0.9 pp, medium confidence |
| Kimi K2.6 | 80.4% | 87.5% | estimated ± 0.9 pp, medium confidence |
| Kimi K3 | 91.3% | 89.7% | estimated ± 0.9 pp, medium confidence |
| Muse Glimmer 30B | 78.8% | 87.2% | estimated ± 0.9 pp, medium confidence |
| Muse Spark | 86.4% | 88.7% | estimated ± 0.9 pp, medium confidence |
| Muse Spark 1.1 | 88.4% | 89.1% | estimated ± 0.9 pp, medium confidence |
| Nemotron 3 Nano Omni 30B A3B | 76.3% | 86.6% | estimated ± 0.9 pp, low confidence |
| Qwen3.5-122B-A10B | 77.2% | 86.8% | estimated ± 0.9 pp, low confidence |
| Qwen3.5 397B | 80.8% | 87.6% | estimated ± 0.9 pp, medium confidence |
| Qwen3.6 Plus | 81.5% | 87.7% | estimated ± 0.9 pp, medium confidence |
| Qwen3.8-27B | 90.2% | 89.5% | estimated ± 0.9 pp, medium confidence |
| Qwen3.8-Flash-Next | 90.6% | 89.5% | estimated ± 0.9 pp, medium confidence |
| Qwen3.8-Omni-Flash | 91.4% | 89.7% | estimated ± 0.9 pp, medium confidence |
| Sakana Fugu | 85.1% | 88.4% | estimated ± 0.9 pp, medium confidence |
| Sakana Fugu-Ultra | 86.6% | 88.7% | estimated ± 0.9 pp, medium confidence |
| Seed 2.1 Pro | 85.4% | 88.5% | estimated ± 0.9 pp, medium confidence |
| Seed 2.1 Turbo | 82.5% | 87.9% | estimated ± 0.9 pp, medium confidence |