benchgap
Agentic · tools

AA Harvey LAB v1.0 leaderboard

As of 2026-11-21, the highest measured score on AA Harvey LAB v1.0 is 94.6% by Kimi K3. 117 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Kimi K394.6%measured
2Kimi K3.1 Pro94.0%estimated ± 1.3 pp, low confidence
3Muse Spark 1.393.7%estimated ± 1.2 pp, medium confidence
4GLM-5.393.6%estimated ± 1.2 pp, medium confidence
5Grok 4.693.6%estimated ± 1.2 pp, medium confidence
6Claude Fable 593.6%measured
7GPT-6 Astra93.5%estimated ± 1.2 pp, medium confidence
8Atria Dawn Preview93.5%estimated ± 9.8 pp, low confidence
9Claude Opus 593.5%measured
10GPT-5.6 Sol93.5%estimated ± 1.2 pp, medium confidence
11Qwen3.8 Max Preview93.4%estimated ± 1.2 pp, medium confidence
12DeepSeek V4 Pro93.4%estimated ± 1.2 pp, medium confidence
13Step 5 Preview93.4%measured
14Ling 3.1 Flash93.3%estimated ± 9.8 pp, low confidence
15Gemini 4 Argon93.3%estimated ± 1.3 pp, medium confidence
16Qwen3.8-27B93.2%estimated ± 1.2 pp, medium confidence
17Claude Sonnet 5.593.1%measured
18Claude Fable 5.193.0%measured
19Inkling 2 Small93.0%estimated ± 9.8 pp, low confidence
20GPT-5.5 Pro92.9%estimated ± 9.8 pp, low confidence
21Claude Sonnet 592.9%estimated ± 1.2 pp, medium confidence
22Muse Spark 1.292.9%estimated ± 1.2 pp, medium confidence
23GPT-5.6 Terra92.9%estimated ± 1.2 pp, medium confidence
24GPT-5.4 Pro92.8%estimated ± 9.8 pp, low confidence
25MiniMax M3.1 Flash92.7%estimated ± 1.3 pp, medium confidence
26GPT-5.6 Luna92.7%estimated ± 1.2 pp, medium confidence
27Claude Opus 4.892.7%estimated ± 1.2 pp, medium confidence
28GPT-6 Sol92.7%estimated ± 1.3 pp, medium confidence
29Grok 4.592.6%estimated ± 1.2 pp, medium confidence
30DeepSeek V4 Flash 073192.5%estimated ± 1.2 pp, medium confidence
31Claude Mythos 592.4%estimated ± 9.8 pp, low confidence
32Gemini 3.8 Flash92.4%estimated ± 1.2 pp, medium confidence
33ERNIE 5.5 Flash Preview92.3%estimated ± 9.8 pp, low confidence
34Claude Opus 4.792.1%estimated ± 1.2 pp, medium confidence
35GLM-5.292.0%estimated ± 1.2 pp, medium confidence
36DeepSeek V4.1 Flash 101792.0%estimated ± 1.7 pp, medium confidence
37LongCat-2.591.8%estimated ± 1.7 pp, medium confidence
38Mistral Large 491.6%estimated ± 9.8 pp, low confidence
39GPT-5.591.4%estimated ± 1.2 pp, medium confidence
40Muse Spark Open91.3%estimated ± 1.7 pp, medium confidence
41Claude Opus 5.591.2%measured
42Hy490.7%estimated ± 1.7 pp, medium confidence
43Gemini 3.7 Flash90.7%measured
44Inkling 2 Mini90.3%estimated ± 9.8 pp, low confidence
45Ling 3.0 Flash Sante90.1%estimated ± 1.7 pp, medium confidence
46Nemotron 3.5 Super VL 120B A12B90.0%estimated ± 9.8 pp, low confidence
47Claude Haiku 5.589.9%measured
48Nemotron 3.5 Super VL 120B A12B NVFP489.9%estimated ± 9.8 pp, low confidence
49Apodex 1.189.8%estimated ± 1.7 pp, medium confidence
50Ornith-1.5-397B89.8%estimated ± 1.7 pp, medium confidence
51Qwen3.8-Flash-Next89.7%estimated ± 1.3 pp, medium confidence
52Qwen3.8 Max89.7%estimated ± 1.3 pp, medium confidence
53Hy4 preview89.6%estimated ± 1.7 pp, medium confidence
54GLM-5.3-Flash89.6%estimated ± 1.7 pp, medium confidence
55Hy4 flash89.5%estimated ± 1.7 pp, medium confidence
56Sage-A1.589.5%estimated ± 1.7 pp, medium confidence
57Quasar 438B89.4%estimated ± 1.2 pp, medium confidence
58Claude Sonnet 4.688.9%estimated ± 1.3 pp, medium confidence
59dots3-note Preview88.7%estimated ± 1.7 pp, medium confidence
60MiniMax M388.4%measured
61Beam 1.0 501B-A23B88.1%estimated ± 1.7 pp, medium confidence
62Beam 1.0 501B-A23B FP888.1%estimated ± 1.7 pp, medium confidence
63Beam 1.0 501B-A23B NVFP488.1%estimated ± 1.7 pp, medium confidence
64Nemotron 3.5 Bolt 74B-A9B BF1688.0%estimated ± 9.8 pp, low confidence
65GLM-587.9%estimated ± 1.7 pp, medium confidence
66Qwen3.5 397B87.8%estimated ± 9.8 pp, low confidence
67Gemini 3.6 Flash87.7%estimated ± 1.2 pp, medium confidence
68Agents-A1-4B87.2%estimated ± 9.8 pp, low confidence
69Nemotron 3.5 Bolt 74B-A9B87.2%estimated ± 9.8 pp, low confidence
70GPT-5.286.9%estimated ± 9.8 pp, low confidence
71Agents-A186.8%estimated ± 1.7 pp, medium confidence
72Step 3.7 Flash86.6%estimated ± 1.7 pp, medium confidence
73Muse Spark 1.185.8%estimated ± 1.2 pp, medium confidence
74Qwen3.5-27B85.7%estimated ± 9.8 pp, low confidence
75Qwen3.5-35B-A3B85.7%estimated ± 9.8 pp, low confidence
76Kimi K2.585.6%estimated ± 9.8 pp, low confidence
77Gemini 3.5 Flash85.6%estimated ± 1.2 pp, medium confidence
78Hy384.3%estimated ± 1.2 pp, medium confidence
79Hy3 Preview84.3%estimated ± 1.2 pp, medium confidence
80GLM-5.184.0%estimated ± 1.2 pp, medium confidence
81Inkling-Small83.8%estimated ± 1.2 pp, medium confidence
82Inkling83.4%estimated ± 1.2 pp, medium confidence
83GLM-4.783.3%estimated ± 9.8 pp, low confidence
84Qwen3.7 Max83.1%estimated ± 1.2 pp, medium confidence
85Sarvam 105B82.7%estimated ± 9.8 pp, low confidence
86Solar Pro 482.6%estimated ± 9.8 pp, low confidence
87LongCat-Flash-Lite-Sparse82.4%estimated ± 9.8 pp, low confidence
88MiMo-V2.5-Pro82.3%estimated ± 1.2 pp, medium confidence
89Kimi K2.7 Code82.2%estimated ± 1.2 pp, medium confidence
90Kimi K2.682.0%estimated ± 1.2 pp, medium confidence
91Nemotron 3 Ultra81.7%measured
92Spark-X2.6-4B81.5%estimated ± 1.7 pp, low confidence
93Ling 3.0 Flash81.4%estimated ± 1.2 pp, low confidence
94Qwen3.6-27B81.0%estimated ± 1.2 pp, low confidence
95Qwen3.7 Plus80.8%estimated ± 1.2 pp, low confidence
96GPT-5.4 mini80.8%estimated ± 1.2 pp, low confidence
97GPT-5.4 nano80.2%estimated ± 1.2 pp, low confidence
98Grok 4.380.1%estimated ± 1.2 pp, low confidence
99MiniMax M2.780.0%estimated ± 1.2 pp, low confidence
100Gemini 3.5 Flash-Lite79.8%estimated ± 1.2 pp, low confidence
101Qwen3.6-35B-A3B79.7%estimated ± 1.2 pp, low confidence
102Muse Glimmer 30B79.4%estimated ± 1.2 pp, low confidence
103Gemini 3.1 Pro79.4%estimated ± 1.2 pp, low confidence
104Qwen3.5-122B-A10B79.4%estimated ± 1.2 pp, low confidence
105Mistral Medium 3.5 128B79.4%estimated ± 1.2 pp, low confidence
106Gemma 4 31B79.4%estimated ± 1.2 pp, low confidence
107GPT-OSS 120B79.4%estimated ± 1.2 pp, low confidence
108Nemotron 3.5 Lightning 30B A3B NVFP479.4%estimated ± 1.2 pp, low confidence
109Granite 4.2 8B79.4%estimated ± 1.2 pp, low confidence
110Command A+79.4%estimated ± 1.2 pp, low confidence
111Gemini 2.5 Pro79.4%estimated ± 1.2 pp, low confidence
112Mistral Large 379.4%estimated ± 1.2 pp, low confidence
113Mistral Small 479.4%estimated ± 1.2 pp, low confidence
114GPT-OSS 20B79.4%estimated ± 1.2 pp, low confidence
115Trinity-Large-Preview79.4%estimated ± 1.2 pp, low confidence
116Trinity-Large-Thinking79.4%estimated ± 1.2 pp, low confidence
117Nemotron 3 Nano 30B79.4%estimated ± 1.2 pp, low confidence
118DeepSeek V379.4%estimated ± 1.2 pp, low confidence
119Celeris-179.4%estimated ± 1.2 pp, low confidence
120Llama 4 Maverick79.4%estimated ± 1.2 pp, low confidence
121Llama 4 Scout79.4%estimated ± 1.2 pp, low confidence
122Gemma 3 27B79.4%estimated ± 1.2 pp, low confidence
123Ornith-1.5-35B-A3B79.2%estimated ± 1.7 pp, low confidence
124Sarvam 30B78.9%estimated ± 9.8 pp, low confidence
125Ornith-1.5-9B77.1%estimated ± 1.7 pp, low confidence
126LFM3-1.7B56.2%estimated ± 1.7 pp, low confidence
127Grok 4.719.6%measured
128GPT-5.40.1%estimated ± 2.0 pp, low confidence
129Claude Opus 4.60.0%estimated ± 2.0 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General