Benchmarks / Agentic

Legal Research Bench

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

77models
–Open Index weight
2026-10-12last updated
Vals AIpublisher

full table below

#ModelEffortScoreCost / taskHarness
1Muse Spark 1.3max55.3%$0.589–
2Opus 5max55.3%$6.76–
3Fable 5.1max55.3%$23.06–
4Gemini 4high54.8%$6.45–
5Wonder Fullmax53.4%$3.60–
6Wondermax50.5%$3.98–
7Opus 5.5max50.5%$25.16–
8Fable 5max49.5%$9.79–
9GLM 5.3max49.0%$2.24–
10Grok 4.6high48.1%$1.53–
11GPT-5.6 Solmax48.1%$21.61–
12Sonnet 5.5max48.1%$14.83–
13Qwen 3.8 Maxmax47.6%$2.49–
14Grok 4.7xhigh47.1%$4.92–
15MiMo V2.6 Prohigh47.1%$0.181–
16Kimi K3max46.2%$3.47–
17GLM 5.3 Flashmax45.2%$0.120–
18Hy4–45.2%$0.846–
19Ember-1–44.7%$3.67–
20Muse Spark 1.2xhigh43.8%$0.508–
21Opus 4.8max43.8%$2.82–
22Haiku 5.5max43.3%$1.63–
23Sonnet 5max41.8%$2.72–
24GLM Legalmax41.8%$0.049–
25DeepSeek V4.1 Flashmax41.3%$0.248–
26GPT-5.6 Terramax41.3%$7.91–
27DeepSeek V4 Pro 0813max40.9%$1.09–
28Step 5 Preview–40.4%$0.839–
29GPT-5.5xhigh40.4%$7.40–
30GPT-6 Astramax39.4%$10.58–
31Gemini 3.8 Flashhigh38.9%$1.63–
32Opus 4.7max38.5%$3.06–
33Sonnet 4.6max38.5%$2.30–
34GPT-6.1 Solmax38.5%$3.38–
35Muse Spark 1.1xhigh38.0%$0.377–
36MiMo V2.6 Flashhigh38.0%$0.070–
37Grok 4.5high38.0%$1.09–
38GPT-5.6 Lunamax36.5%$0.845–
39Qwen 3.8 27Bxhigh36.1%$2.16–
40Gemini 3.7 Flashhigh34.6%$0.760–
41Mistral Large 4high31.7%$5.33–
42GLM 5.2max31.3%$0.889–
43Gemini 3.5 Flashhigh30.8%$1.31–
44DeepSeek V4 Flash 0731high30.3%$0.275–
45GPT-6 Lunamax30.3%$0.440–
46MiniMax M3max29.8%$0.343–
47GPT-6 Solmax28.8%$4.90–
48Inklingmax28.4%$0.494–
49GLM 5.1max27.9%$0.386–
50Inkling Smallxhigh25.5%$0.142–
51Qwen 3.7 Maxmax25.5%$1.40–
52Gemini 3.6 Flashhigh25.0%$0.467–
53DeepSeek V4max23.1%$0.682–
54Gemini 3.1 Prohigh20.7%$1.10–
55Gemini 3 Flashhigh18.3%$0.135–
56Qwen 3.7 Plushigh16.3%$0.296–
57MiMo V2.5 Proxhigh15.9%$0.083–
58Kimi K2.5max15.9%$0.230–
59Kimi K2.6max15.9%$1.52–
60Grok 4.3high15.4%$0.203–
61Ling 3.0 Flash Fin–15.4%$0.028–
62Nemotron 3 Ultra 550B A55Bmax15.4%––
63Qwen 3.6 Plusmax14.9%$0.698–
64Gemini 3.5 Flash-Litehigh13.9%$0.125–
65Grok 4.20xhigh13.9%$0.261–
66GPT-5.4 Minixhigh12.5%$1.86–
67Haiku 4.5xhigh10.6%$0.269–
68MiniMax M2.7high10.6%$0.219–
69MiMo V2.5xhigh9.1%$0.033–
70Mistral Medium 3.5high9.1%$1.31–
71GPT-5.4 Nanohigh6.3%$0.141–
72Mercury 2.5high4.3%$0.058–
73Gemini 3.1 Flash-Litehigh3.4%$0.014–
74Nemotron 3.5 Lightning–2.9%$0.0044–
75Laguna M.1–2.4%$0.220–
76Laguna XS.2–1.0%$0.108–
77Ling 3.0 Flash–0.0%$0.036–