FinFIRST - Computation and Answer Formation: leaderboard
The computation and answer formation rubric of FinFIRST (Wang et al., 2026): the percentage of rubric points for calculation, reasoning and an evidence-supported final answer in expert-authored financial research questions. Every model runs in the same ReAct harness with web search, page visit and Python tools, and a GLM-5.1 judge checks each criterion; higher is better.
Metric: Weighted rubric score (%; the 2,893 rubric points for calculation, reasoning and an evidence-supported final answer; 123 expert-authored bilingual financial research tasks, each scored against expert atomic rubric criteria (701 in all, weights summing to 100 per task) by a GLM-5.1 judge validated against finance experts (Cohen kappa 0.816); all models run in the same ReAct harness with web search, page visit and Python tools, temperature 1.0, one run; missing outputs count as failed). Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 82.72 | #20 |
| 2 | GPT-5.6 Sol | 81.58 | #29 |
| 3 | Kimi K3 | 77.98 | #53 |
| 4 | GLM-5.3 | 75.77 | #62 |
| 5 | Gemini 3.7 Flash | 72.8 | #32 |
| 6 | Qwen 3.8 Flash | 71.93 | #109 |
| 7 | Qwen 3.8 Max | 69.72 | #55 |
| 8 | Qwen 3.8 27B | 66.92 | #199 |
| 9 | GLM-5.3 Flash | 66.44 | #119 |
| 10 | DeepSeek V4 Pro (0813) | 64.88 | #81 |
| 11 | Ling-3.0-flash-fin | 61.25 | #289 |
| 12 | DeepSeek V4 Flash (0731) | 55.69 | #136 |
| 13 | GLM-5.2 | 54.37 | #137 |
| 14 | MiniMax-M3 | 46.87 | #265 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=finfirst-computation-and-answer-formation · How It Works · Data refreshed daily, snapshot 2026-11-12.