FrontierFinance: leaderboard
FrontierFinance (Samaya AI, 2026) poses 220 open-ended financial research queries across data and modeling, sector and macro analysis, earnings, company research, catalyst monitoring and screening. Each response is checked against expert-written rubric items by a three-judge panel (Claude Sonnet 4.6, Gemini 3.1 Pro and GPT-5.4, majority vote), and the score is the mean per-query share of rubric items satisfied (%); higher is better. Models run in the Finance Agent v2 harness at the reasoning effort the benchmark site lists.
Metric: Rubric qualification rate (%, macro average over queries, three-judge majority vote). Source: research.samaya.ai. Saturation forecast: Around 2029. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.1 (High) | 55.86 | #14 (Claude Fable 5.1) |
| 2 | Muse Spark Open (xHigh) | 53.73 | #79 (Muse Spark Open) |
| 3 | Muse Wonder | 52.56 | #27 |
| 4 | Claude Opus 5.5 (High) | 50.88 | #5 (Claude Opus 5.5) |
| 5 | Gemini 3.8 Flash (Medium) | 50.01 | #36 (Gemini 3.8 Flash) |
| 6 | Claude Fable 5 (High) | 49.19 | #26 (Claude Fable 5) |
| 7 | GPT-5.6 Sol (Medium) | 46.83 | #29 (GPT-5.6 Sol) |
| 8 | Ling-3.0-flash-fin | 46.59 | #289 |
| 9 | Kimi K3 (Max) | 46.37 | #53 (Kimi K3) |
| 10 | Gemini 3.6 Flash (Medium) | 46.27 | #72 (Gemini 3.6 Flash) |
| 11 | Claude Opus 4.8 (High) | 45.04 | #56 (Claude Opus 4.8) |
| 12 | GPT-5.5 (Medium) | 43.47 | #48 (GPT-5.5) |
| 13 | GLM-5.2 (Max) | 42.77 | #137 (GLM-5.2) |
| 14 | DeepSeek V4 Pro (High) | 40.54 | #181 (DeepSeek V4 Pro) |
| 15 | Kimi K2.6 | 32.32 | #173 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #6 GPT-6 Astra.
Interactive version: theaggregate.ai/benchmark?slug=frontierfinance · How It Works · Data refreshed daily, snapshot 2026-11-12.