Calibration
CharXiv → ScreenSpot Pro
ScreenSpot Pro is estimated from CharXiv with a Hill curve fitted on 13 models measured on both: y = 0.0000 + (1.0714 − 0.0000)·x^6.00 / (0.71072^6.00 + x^6.00), R² = 0.76, cross-validated error 6.8 pp. It is used for 29 estimates.
| Estimated model | CharXiv | ScreenSpot Pro | Source |
|---|---|---|---|
| Claude Mythos 5 | 93.5% | 89.8% | estimated ± 6.8 pp, medium confidence |
| Kimi K3.1 Pro | 93.4% | 89.7% | estimated ± 6.8 pp, medium confidence |
| Glimpse | 92.9% | 89.3% | estimated ± 6.8 pp, medium confidence |
| Kimi K3 | 91.3% | 87.6% | estimated ± 6.8 pp, medium confidence |
| Claude Opus 4.7 | 91.0% | 87.3% | estimated ± 6.8 pp, medium confidence |
| LongCat-2.5 | 90.2% | 86.5% | estimated ± 6.8 pp, medium confidence |
| GLM-5.3-Flash | 89.4% | 85.5% | estimated ± 6.8 pp, medium confidence |
| Gemini 3.7 Flash | 88.7% | 84.7% | estimated ± 6.8 pp, medium confidence |
| Muse Spark 1.1 | 88.4% | 84.4% | estimated ± 6.8 pp, medium confidence |
| Claude Sonnet 5 | 88.3% | 84.2% | estimated ± 6.8 pp, medium confidence |
| Sakana Fugu-Ultra | 86.6% | 82.1% | estimated ± 6.8 pp, medium confidence |
| Seed 2.1 Pro | 85.4% | 80.4% | estimated ± 6.8 pp, medium confidence |
| Sakana Fugu | 85.1% | 80.0% | estimated ± 6.8 pp, medium confidence |
| Gemini 3.5 Flash | 84.2% | 78.7% | estimated ± 6.8 pp, medium confidence |
| MiniMax M3.1 Flash | 82.5% | 76.1% | estimated ± 6.8 pp, medium confidence |
| Seed 2.1 Turbo | 82.5% | 76.1% | estimated ± 6.8 pp, medium confidence |
| Inkling | 82.0% | 75.2% | estimated ± 6.8 pp, medium confidence |
| Inkling-Small | 81.3% | 74.1% | estimated ± 6.8 pp, medium confidence |
| MiMo-V2.5 | 81.0% | 73.6% | estimated ± 6.8 pp, medium confidence |
| Nemotron 3.5 Super VL 120B A12B | 80.8% | 73.3% | estimated ± 6.8 pp, medium confidence |
| Kimi K2.6 | 80.4% | 72.5% | estimated ± 6.8 pp, medium confidence |
| Nemotron 3.5 Super VL 120B A12B NVFP4 | 79.9% | 71.6% | estimated ± 6.8 pp, medium confidence |
| Qwen3.5-27B | 79.5% | 70.9% | estimated ± 6.8 pp, medium confidence |
| Qwen3.5-35B-A3B | 77.5% | 67.2% | estimated ± 6.8 pp, medium confidence |
| Claude Sonnet 4.6 | 77.4% | 67.0% | estimated ± 6.8 pp, medium confidence |
| Qwen3.5-122B-A10B | 77.2% | 66.6% | estimated ± 6.8 pp, medium confidence |
| Gemini 3.1 Flash-Lite | 73.2% | 58.3% | estimated ± 6.8 pp, medium confidence |
| Grok 4.20 | 60.9% | 30.4% | estimated ± 6.8 pp, low confidence |
| Command A+ | 52.7% | 15.3% | estimated ± 6.8 pp, low confidence |
Fitted on
| Model | CharXiv | ScreenSpot Pro |
|---|---|---|
| Claude Opus 4.5 | 68.5% | 45.7% |
| Claude Opus 4.8 | 89.9% | 87.9% |
| Gemini 3.1 Pro | 80.2% | 84.4% |
| Gemini 3 Pro | 81.4% | 72.7% |
| GPT-5.4 | 82.8% | 85.4% |
| Inkling 2 Small | 87.7% | 83.8% |
| Muse Glimmer 30B | 78.8% | 75.4% |
| Muse Spark | 86.4% | 84.1% |
| Nemotron 3 Nano Omni 30B A3B | 76.3% | 57.8% |
| Qwen3.5 397B | 80.8% | 65.6% |
| Qwen3.6 Plus | 81.5% | 68.2% |
| Qwen3.7 Plus | 85.9% | 79.0% |
| Qwen3.8 Max | 93.5% | 84.5% |