ATLAS-Finance v1.1 Pass@3
Professional Work · 2026-10-15
Share of ATLAS-Finance v1.1.0's 100 investment-banking tasks passed in at least one of three rollouts, under the same pass rule (every criterion weighted 3 or more met, no penalty triggered). It is a component of the v1.1 factor group, computed from the same graded rollouts as the headline pass rate.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 15.0 |
| GPT-6 Astra | 14.0 |
| Claude Fable 5.1 | 13.0 |
| GLM-5.3 | 7.0 |
| GPT-5.6 Sol | 6.0 |
| Grok 4.6 | 6.0 |
| Gemini 3.7 Flash | 5.0 |
| Kimi K3 | 5.0 |
| Gemini 3.8 Flash | 4.0 |
| Gemini 3.1 Pro Preview | 2.0 |