# Vals Legal Research Bench (Legal Research Bench)

> Evaluating agents on legal research tasks across diverse areas of US law

Canonical page: https://benchlm.ai/benchmarks/legalresearchbench

- Category: [external](/external)
- Last updated: September 10, 2026

## About Legal Research Bench

- Year: 2026
- Tasks: US-law legal research tasks
- Format: Accuracy score
- Difficulty: Professional legal research
- Paper: [Legal Research Bench](https://www.vals.ai/benchmarks/legal_research)

BenchLM mirrors the public Vals AI Legal Research Bench leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

Legal Research Bench is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (59 models)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [Muse Spark 1.3 Max](https://www.vals.ai/models/meta_muse_spark_1_3_max) | max reasoning | Meta | 55.29% |
| 2 | [Claude Opus 5](/models/claude-opus-5) | — | Anthropic | 55.29% |
| 3 | [Claude Fable 5.1](/models/claude-fable-5-1) | — | Anthropic | 55.29% |
| 4 | [Claude Fable 5](/models/claude-fable) | — | Anthropic | 49.52% |
| 5 | [GLM-5.3](/models/glm-5-3) | max reasoning | Z.AI | 49.04% |
| 6 | [Grok 4.6](/models/grok-4-6) | high reasoning | xAI | 48.08% |
| 7 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | max reasoning | OpenAI | 48.08% |
| 8 | [Qwen3.8 Max](/models/qwen3-8-max) | — | Alibaba | 47.60% |
| 9 | [GLM-5.3-Flash](/models/glm-5-3-flash) | max reasoning | Z.AI | 45.19% |
| 10 | [Kimi K3](/models/kimi-k3) | — | Moonshot AI | 44.23% |
| 11 | [Muse Spark 1.2](/models/muse-spark-1-2) | xhigh reasoning | Meta | 43.75% |
| 12 | [Claude Opus 4.8](/models/claude-opus-4-8) | — | Anthropic | 43.75% |
| 13 | [Claude Sonnet 5](/models/claude-sonnet-5) | — | Anthropic | 41.83% |
| 14 | [DeepSeek V4.1 Flash](/models/deepseek-v4-1-flash) | high reasoning | DeepSeek | 41.35% |
| 15 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | max reasoning | OpenAI | 41.35% |
| 16 | [Muse Spark 1.3](/models/muse-spark-1-3) | xhigh reasoning | Meta | 40.87% |
| 17 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | max reasoning | DeepSeek | 40.87% |
| 18 | [GPT-5.5](/models/gpt-5-5) | xhigh reasoning | OpenAI | 40.38% |
| 19 | [GPT-6 Astra](/models/gpt-6-astra) | max reasoning | OpenAI | 39.42% |
| 20 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | high reasoning | Google | 38.94% |
| 21 | [Claude Opus 4.7](/models/claude-opus-4-7) | — | Anthropic | 38.46% |
| 22 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | — | Anthropic | 38.46% |
| 23 | [Muse Spark 1.1](/models/muse-spark-1-1) | xhigh reasoning | Meta | 37.98% |
| 24 | [Grok 4.5](/models/grok-4-5) | high reasoning | xAI | 37.98% |
| 25 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | max reasoning | OpenAI | 36.54% |
| 26 | [Qwen3.8-27B](/models/qwen3-8-27b) | xhigh reasoning | Alibaba | 36.06% |
| 27 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | high reasoning | Google | 34.62% |
| 28 | [GLM-5.2](/models/glm-5-2) | max reasoning | Z.AI | 31.25% |
| 29 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | high reasoning | Google | 30.77% |
| 30 | [DeepSeek V4 Flash 0731](/models/deepseek-v4-flash-0731) | high reasoning | DeepSeek | 30.29% |
| 31 | [MiniMax M3](/models/minimax-m3) | — | MiniMax | 29.81% |
| 32 | [Inkling](/models/inkling) | 0.99 reasoning | Thinking Machines Lab | 28.36% |
| 33 | [GLM-5.1](/models/glm-5-1) | — | Z.AI | 27.89% |
| 34 | [Inkling-Small](/models/inkling-small) | 0.99 reasoning | Thinking Machines Lab | 25.48% |
| 35 | [Qwen3.7 Max](/models/qwen3-7-max) | — | Alibaba | 25.48% |
| 36 | [Gemini 3.6 Flash](/models/gemini-3-6-flash) | high reasoning | Google | 25.00% |
| 37 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | max reasoning | DeepSeek | 23.08% |
| 38 | [Gemini 3.1 Pro Preview](https://www.vals.ai/models/google_gemini-3.1-pro-preview) | high reasoning | Google | 20.67% |
| 39 | [Gemini 3 Flash Preview](https://www.vals.ai/models/google_gemini-3-flash-preview) | high reasoning | Google | 18.27% |
| 40 | [Qwen3.7 Plus](/models/qwen3-7-plus) | — | Alibaba | 16.35% |
| 41 | [MiMo-V2.5-Pro](/models/mimo-v2-5-pro) | — | Xiaomi | 15.87% |
| 42 | [Kimi K2.5 Thinking](https://www.vals.ai/models/kimi_kimi-k2.5-thinking) | — | Moonshot AI | 15.87% |
| 43 | [Kimi K2.6](/models/kimi-2-6) | — | Moonshot AI | 15.87% |
| 44 | [Grok 4.3](/models/grok-4-3) | high reasoning | xAI | 15.38% |
| 45 | [Nemotron 3 Ultra 550b A55b](https://www.vals.ai/models/nvidia_nemotron-3-ultra-550b-a55b) | — | Nvidia | 15.38% |
| 46 | [Qwen3.6 Plus](/models/qwen3-6-plus) | — | Alibaba | 14.90% |
| 47 | [Gemini 3.5 Flash-Lite](/models/gemini-3-5-flash-lite) | high reasoning | Google | 13.94% |
| 48 | [Grok 4.20 0309 Reasoning](https://www.vals.ai/models/grok_grok-4.20-0309-reasoning) | — | xAI | 13.94% |
| 49 | [GPT-5.4 mini](/models/gpt-5-4-mini) | xhigh reasoning | OpenAI | 12.50% |
| 50 | [Claude Haiku 4.5 Thinking](/models/claude-haiku-4-5-thinking) | — | Anthropic | 10.58% |
| 51 | [MiniMax M2.7](/models/minimax-m2-7) | — | MiniMax | 10.58% |
| 52 | [MiMo-V2.5](/models/mimo-v2-5) | — | Xiaomi | 9.13% |
| 53 | [Mistral Medium 3.5](https://www.vals.ai/models/mistralai_mistral-medium-3.5) | high reasoning | Mistral AI | 9.13% |
| 54 | [GPT-5.4 nano](/models/gpt-5-4-nano) | high reasoning | OpenAI | 6.25% |
| 55 | [Gemini 3.1 Flash Lite Preview](https://www.vals.ai/models/google_gemini-3.1-flash-lite-preview) | — | Google | 3.37% |
| 56 | [Nemotron Lightning 3p5 30b A3b](https://www.vals.ai/models/fireworks_nemotron-lightning-3p5-30b-a3b) | — | Fireworks AI | 2.88% |
| 57 | [Laguna M.1](/models/laguna-m-1) | — | Poolside | 2.40% |
| 58 | [Laguna XS.2](/models/laguna-xs-2) | — | Poolside | 0.96% |
| 59 | [Ling 3.0 Flash 2607](https://www.vals.ai/models/ant_ling-3.0-flash-2607) | — | Ant | 0.00% |

## FAQ

### What does Legal Research Bench measure?

Evaluating agents on legal research tasks across diverse areas of US law

### Which model leads the published Legal Research Bench snapshot?

Muse Spark 1.3 Max currently leads the published Legal Research Bench snapshot with a score of 55.29%.

### How many models are evaluated on Legal Research Bench?

The September 10, 2026 contains 59 AI models.

### Does Legal Research Bench affect BenchLM's overall score?

Not directly. Legal Research Bench is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.