FrontierFinance v3.1
Professional Work · 2026-07-06
Samaya AI's FrontierFinance v3.1 evaluates finance agents on 220 open-ended, dated research queries across six investor workflows, using 11,543 expert-authored rubric criteria. The headline metric is rubric qualification rate macro-averaged across queries after three-model majority-vote judging.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 49.2 |
| GPT-5.6 Sol | 46.8 |
| Opus 4.8 | 45.0 |
| GPT-5.5 | 43.5 |
| GLM-5.2 | 42.8 |
| DeepSeek-V4-Pro | 40.5 |
| Kimi K2.6 | 32.3 |
| Gemini 3.1 Pro Preview | 30.7 |