# Vals Finance Agent v2.1 (Finance Agent v2.1)

> Financial research with tools, revised questions and rubrics, a new judge jury, and stricter treatment of hedged answers.

Canonical page: https://benchlm.ai/benchmarks/finance-agent-v2-1

- Category: [Agentic](/agentic)
- Last updated: October 27, 2026

Evaluation results published by [Vals AI](https://www.vals.ai/benchmarks/fabv2-1). Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.

## About Finance Agent v2.1

- Year: 2026
- Tasks: 449 private test questions across nine analyst categories
- Format: Dealbreaker-gated partial credit; all-pass reported separately
- Difficulty: Professional expert-task agent workflow
- Paper: [Finance Agent v2.1](https://www.vals.ai/benchmarks/fabv2-1)

The primary metric awards severity-weighted partial credit after all dealbreakers pass. All-pass requires every check. Version 2.1 revises the jury, data, and hedging rules; Vals says scores are not comparable with v2. We retain both editions as separate display-only tables.

Finance Agent v2.1 is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (33 models)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [Claude Opus 5.5](/models/claude-opus-5-5) | — | Anthropic | 67.85% |
| 2 | [Gemini 4 Argon](/models/gemini-4-argon) | high reasoning | Google | 67.66% |
| 3 | [Muse Wonder Full](/models/muse-wonder-full) | max reasoning | Meta | 67.58% |
| 4 | [Claude Sonnet 5.5](/models/claude-sonnet-5-5) | — | Anthropic | 67.28% |
| 5 | [Gemini 4 Carbon](/models/gemini-4-carbon) | high reasoning | Google | 67.16% |
| 6 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | high reasoning | Google | 65.76% |
| 7 | [Muse Wonder](/models/muse-wonder) | max reasoning | Meta | 65.55% |
| 8 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | max reasoning | DeepSeek | 65.43% |
| 9 | [Claude Haiku 5.5](/models/claude-haiku-5-5) | — | Anthropic | 65.37% |
| 10 | [GLM-5.3](/models/glm-5-3) | max reasoning | Z.AI | 65.00% |
| 11 | [MiMo-V2.6-Pro](/models/mimo-v2-6-pro) | — | Xiaomi | 64.26% |
| 12 | [MiMo-V2.6-Flash](/models/mimo-v2-6-flash) | — | Xiaomi | 63.32% |
| 13 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | max reasoning | DeepSeek | 62.93% |
| 14 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | max reasoning | OpenAI | 61.82% |
| 15 | [Hy4 preview](/models/hy4-preview) | — | Tencent | 61.62% |
| 16 | [Kimi K3.1 Pro](/models/kimi-3-1-pro) | max reasoning | Moonshot AI | 61.49% |
| 17 | [Mistral Large 4](/models/mistral-large-4) | high reasoning | Mistral | 61.38% |
| 18 | [Muse Spark Open](/models/muse-spark-open) | xhigh reasoning | Meta | 60.66% |
| 19 | [GLM-5.3-Flash](/models/glm-5-3-flash) | max reasoning | Z.AI | 60.59% |
| 20 | [Grok 4.7](/models/grok-4-7) | xhigh reasoning | xAI | 60.59% |
| 21 | [GPT-6 Astra](/models/gpt-6-astra) | max reasoning | OpenAI | 60.43% |
| 22 | [Kimi K3](/models/kimi-k3) | max reasoning | Moonshot AI | 60.01% |
| 23 | [Step 5 Preview](/models/step-5-preview) | — | StepFun | 58.90% |
| 24 | [Qwen3.8 Max Internal](https://www.vals.ai/models/alibaba_qwen3.8-max-internal) | — | Alibaba | 58.83% |
| 25 | [MiniMax M3.1 Flash](/models/minimax-m3-1-flash) | — | MiniMax | 58.83% |
| 26 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | max reasoning | DeepSeek | 57.12% |
| 27 | [GPT-6 Luna](/models/gpt-6-luna) | max reasoning | OpenAI | 56.84% |
| 28 | [GPT-6.1 Terra](/models/gpt-6-1-terra) | max reasoning | OpenAI | 55.96% |
| 29 | [Mercury 3](/models/mercury-3) | high reasoning | Inception | 54.71% |
| 30 | [MiniMax M3](/models/minimax-m3) | — | MiniMax | 54.34% |
| 31 | [Beam 1.0](https://www.vals.ai/models/reflection_Beam-1.0) | — | Reflection | 51.83% |
| 32 | [Inkling](/models/inkling) | 0.99 reasoning | Thinking Machines Lab | 49.87% |
| 33 | [Inkling-Small](/models/inkling-small) | 0.99 reasoning | Thinking Machines Lab | 44.78% |

## FAQ

### What does Finance Agent v2.1 measure?

Financial research with tools, revised questions and rubrics, a new judge jury, and stricter treatment of hedged answers.

### Which model leads the published Finance Agent v2.1 snapshot?

Claude Opus 5.5 currently leads the published Finance Agent v2.1 snapshot with a score of 67.85%.

### How many models are evaluated on Finance Agent v2.1?

The October 27, 2026 contains 33 AI models.

### Does Finance Agent v2.1 affect BenchLM's overall score?

Not directly. Finance Agent v2.1 is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.