Calibration
JobBench → DeepSearchQA
DeepSearchQA is estimated from JobBench with a linear curve fitted on 8 models measured on both: y = 0.3148·x + 0.7021, R² = 0.44, cross-validated error 8.8 pp. It is used for 15 estimates.
| Estimated model | JobBench | DeepSearchQA | Source |
|---|---|---|---|
| Muse Wonder Full | 75.3% | 93.9% | estimated ± 8.8 pp, low confidence |
| Muse Wonder | 70.5% | 92.4% | estimated ± 8.8 pp, low confidence |
| Ultramarine 1 Flash | 62.9% | 90.0% | estimated ± 8.8 pp, low confidence |
| Muse Spark Open | 61.8% | 89.7% | estimated ± 8.8 pp, low confidence |
| Hy4 preview | 61.7% | 89.6% | estimated ± 8.8 pp, low confidence |
| Qwen3.8-Flash-Next | 55.7% | 87.8% | estimated ± 8.8 pp, low confidence |
| GPT-5.1-Codex | 26.2% | 78.5% | estimated ± 8.8 pp, low confidence |
| GPT-5.2-Codex | 26.0% | 78.4% | estimated ± 8.8 pp, low confidence |
| Claude 4.1 Opus | 21.9% | 77.1% | estimated ± 8.8 pp, low confidence |
| Qwen3.5 Plus | 18.5% | 76.0% | estimated ± 8.8 pp, low confidence |
| Claude 4 Sonnet | 18.4% | 76.0% | estimated ± 8.8 pp, low confidence |
| Claude Haiku 4.5 | 16.0% | 75.3% | estimated ± 8.8 pp, low confidence |
| Gemini 3 Flash | 11.4% | 73.8% | estimated ± 8.8 pp, low confidence |
| Gemini 3 Pro | 11.4% | 73.8% | estimated ± 8.8 pp, low confidence |
| GPT-5 | 8.5% | 72.9% | estimated ± 8.8 pp, low confidence |
Fitted on
| Model | JobBench | DeepSearchQA |
|---|---|---|
| Atria Dawn Preview | 50.3% | 96.0% |
| Claude Opus 4.6 | 36.7% | 73.7% |
| GPT-5.4 | 38.9% | 73.6% |
| Inkling 2 Small | 70.5% | 90.9% |
| Kimi K2.5 | 8.7% | 77.1% |
| Kimi K3 | 52.9% | 95.0% |
| Muse Spark 1.1 | 54.7% | 84.9% |
| Muse Spark 1.3 | 64.9% | 89.4% |