FinFIRST - Computation and Answer Formation: leaderboard
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 December 2026. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 82.72 |
| 2 | GPT-5.6 Sol | 81.58 |
| 3 | Kimi K3 | 77.98 |
| 4 | GLM-5.3 | 75.77 |
| 5 | Gemini 3.7 Flash | 72.8 |
| 6 | Qwen 3.8 Flash | 71.93 |
| 7 | Qwen 3.8 Max | 69.72 |
| 8 | Qwen 3.8 27B | 66.92 |
| 9 | GLM-5.3 Flash | 66.44 |
| 10 | DeepSeek V4 Pro (0813) | 64.88 |
| 11 | Ling-3.0-flash-fin | 61.25 |
| 12 | DeepSeek V4 Flash (0731) | 55.69 |
| 13 | GLM-5.2 | 54.37 |
| 14 | MiniMax-M3 | 46.87 |
Interactive version: theaggregate.ai/benchmark?slug=finfirst-computation-and-answer-formation · How It Works · Data refreshed daily, snapshot 2026-09-26.