FrontierFinance v3.1 — Screening and Discovery
Professional Work · 2026-07-06
Screening and Discovery use-case slice of Samaya AI's FrontierFinance v3.1, containing 17 of the 220 public queries. It reports rubric qualification rate macro-averaged across queries in the slice.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 39.9 |
| Kimi K3 | 36.9 |
| GPT-5.6 Sol | 36.5 |
| Gemini 3.6 Flash | 36.1 |
| Gemini 3.8 Flash | 35.4 |
| GPT-5.5 | 34.8 |
| Claude Fable 5 | 33.3 |
| Ling 3.0 Flash Fin | 31.8 |
| Opus 4.8 | 30.2 |
| DeepSeek-V4-Pro | 28.1 |
| Kimi K2.6 | 27.8 |
| Gemini 3.1 Pro Preview | 27.2 |
| GLM-5.2 | 24.9 |