benchgap
Knowledge & reasoning

LABBench2 leaderboard

As of 2026-11-21, the highest measured score on LABBench2 is 88.8% by Gemini 4 Argon. 113 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.290.7%estimated ± 2.5 pp, low confidence
2Gemini 4 Argon88.8%measured
3Gemini 3.8 Flash86.2%measured
4Claude Sonnet 5.586.2%estimated ± 2.4 pp, medium confidence
5Claude Fable 586.1%estimated ± 1.7 pp, medium confidence
6GPT-6.2 Astra86.1%estimated ± 1.7 pp, medium confidence
7Muse Wonder Full85.6%estimated ± 2.4 pp, medium confidence
8Kimi K3.1 Pro85.1%estimated ± 2.4 pp, medium confidence
9Claude Fable 5.184.6%estimated ± 1.7 pp, medium confidence
10Claude Opus 5.584.6%estimated ± 1.7 pp, medium confidence
11Gemini 4 Carbon84.6%estimated ± 1.7 pp, medium confidence
12GPT-6.2 Sol84.6%estimated ± 1.7 pp, medium confidence
13GPT-6 Astra84.6%estimated ± 1.7 pp, medium confidence
14Claude Opus 584.2%measured
15Muse Wonder84.1%estimated ± 2.4 pp, medium confidence
16MiMo-V2.6-Pro83.8%estimated ± 2.4 pp, medium confidence
17Gemini 3.1 Pro83.6%estimated ± 1.8 pp, low confidence
18MiniMax M3.1 Flash83.2%estimated ± 2.4 pp, medium confidence
19Gemini 3 Pro83.1%estimated ± 1.8 pp, low confidence
20GPT-6.1 Sol83.1%estimated ± 1.7 pp, medium confidence
21Inkling 2 Small83.1%estimated ± 1.7 pp, medium confidence
22Claude Opus 4.783.0%estimated ± 1.8 pp, low confidence
23Grok 4.682.8%estimated ± 1.8 pp, low confidence
24Gemini 3.6 Flash82.7%estimated ± 1.8 pp, low confidence
25Qwen3.7 Max82.7%estimated ± 1.8 pp, low confidence
26Muse Spark 1.382.7%estimated ± 1.8 pp, low confidence
27Grok 4.582.7%estimated ± 1.8 pp, low confidence
28Step 5 Preview82.6%estimated ± 2.4 pp, medium confidence
29GPT-6.1 Terra82.4%estimated ± 2.4 pp, medium confidence
30Muse Spark 1.182.4%estimated ± 1.8 pp, low confidence
31Gemini 3 Flash82.3%estimated ± 1.8 pp, low confidence
32Qwen3.8 Max82.3%estimated ± 1.8 pp, low confidence
33Muse Spark 1.282.2%estimated ± 1.8 pp, low confidence
34Gemini 3.7 Flash82.1%measured
35GPT-5.6 Sol82.1%measured
36Claude Opus 4.682.0%estimated ± 2.5 pp, low confidence
37Qwen3.6 Plus81.8%estimated ± 1.8 pp, low confidence
38Kimi K2.681.8%estimated ± 1.8 pp, low confidence
39Claude Sonnet 4.5 Thinking81.7%estimated ± 1.8 pp, low confidence
40Claude Haiku 5.581.7%estimated ± 2.4 pp, medium confidence
41Claude Opus 4.5 Thinking81.6%estimated ± 1.8 pp, low confidence
42Claude Sonnet 4.681.6%estimated ± 1.8 pp, low confidence
43Muse Spark81.6%estimated ± 1.8 pp, low confidence
44Qwen3.8 Max Preview81.5%estimated ± 1.8 pp, low confidence
45GPT-6 Sol81.5%estimated ± 1.7 pp, medium confidence
46DeepSeek V4 Pro81.5%estimated ± 1.8 pp, low confidence
47GLM-5.181.4%estimated ± 1.8 pp, low confidence
48GLM-5.381.3%estimated ± 1.8 pp, low confidence
49GLM-5.281.3%estimated ± 1.8 pp, low confidence
50dots3-note Preview81.2%estimated ± 3.1 pp, low confidence
51GPT-5.6 Terra81.2%measured
52GPT-581.2%estimated ± 1.8 pp, low confidence
53GPT-5.181.1%estimated ± 1.8 pp, low confidence
54Grok 4.781.1%estimated ± 2.4 pp, medium confidence
55Grok 4.2081.1%estimated ± 1.8 pp, low confidence
56Inkling81.1%estimated ± 1.8 pp, low confidence
57DeepSeek V4 Flash 073181.0%estimated ± 1.8 pp, low confidence
58Gemini 3.1 Flash-Lite81.0%estimated ± 1.8 pp, low confidence
59GPT-5.281.0%estimated ± 1.8 pp, low confidence
60GLM-5.3-Flash80.9%estimated ± 1.8 pp, low confidence
61GPT-5.6 Luna80.9%estimated ± 1.8 pp, low confidence
62Gemini 3.5 Flash-Lite80.8%estimated ± 1.8 pp, low confidence
63Grok 4.380.8%estimated ± 1.8 pp, low confidence
64Nemotron 3 Ultra80.8%estimated ± 1.8 pp, low confidence
65Qwen3.8-Flash-Next80.7%estimated ± 1.8 pp, low confidence
66Claude Opus 4.580.7%estimated ± 1.8 pp, low confidence
67Inkling-Small80.7%estimated ± 1.8 pp, low confidence
68o380.7%estimated ± 1.8 pp, low confidence
69GPT-5.580.6%estimated ± 1.7 pp, low confidence
70GPT-5.5 Pro80.6%estimated ± 1.7 pp, low confidence
71Grok 480.5%estimated ± 1.8 pp, low confidence
72GPT-5.4 mini80.1%estimated ± 1.8 pp, low confidence
73MiMo-V2.5-Pro80.1%estimated ± 1.8 pp, low confidence
74Claude Sonnet 580.1%measured
75Qwen3 Max80.0%estimated ± 1.8 pp, low confidence
76Qwen3.8-27B79.9%estimated ± 1.8 pp, low confidence
77Grok 4.1 Fast (Reasoning)79.9%estimated ± 1.8 pp, low confidence
78MiniMax M379.9%estimated ± 1.8 pp, low confidence
79Qwen3.5 Flash79.8%estimated ± 1.8 pp, low confidence
80GPT-5.4 Pro79.8%estimated ± 1.7 pp, low confidence
81Kimi K379.8%estimated ± 1.7 pp, low confidence
82Ling 3.1 Flash79.5%estimated ± 2.4 pp, low confidence
83o179.5%estimated ± 1.8 pp, low confidence
84DeepSeek V4.1 Flash 101779.5%estimated ± 2.4 pp, low confidence
85GPT-5.3 Codex79.4%estimated ± 1.8 pp, low confidence
86GPT-6 Luna79.2%estimated ± 2.4 pp, low confidence
87MiMo-V2.579.1%estimated ± 1.8 pp, low confidence
88Claude 3.7 Sonnet Thinking79.0%estimated ± 1.8 pp, low confidence
89GLM-4.779.0%estimated ± 1.8 pp, low confidence
90GLM-4.678.7%estimated ± 1.8 pp, low confidence
91Grok 4 Fast (Reasoning)78.7%estimated ± 1.9 pp, low confidence
92Ling 3.0 Flash78.6%estimated ± 1.8 pp, low confidence
93GPT-5.478.4%estimated ± 1.7 pp, low confidence
94MiMo-V2.6-Flash77.7%estimated ± 2.4 pp, low confidence
95Beam 1.0 501B-A23B77.5%estimated ± 2.4 pp, low confidence
96Apodex 1.177.3%estimated ± 2.4 pp, low confidence
97Apodex 1.1 Mini77.3%estimated ± 2.4 pp, low confidence
98Claude Opus 4.6 (Adaptive)77.2%estimated ± 1.7 pp, low confidence
99Qwen3.7 Plus77.1%estimated ± 1.8 pp, low confidence
100GPT-5.2-Codex76.9%estimated ± 1.8 pp, low confidence
101Hy376.6%estimated ± 1.8 pp, low confidence
102Hy3 Preview76.6%estimated ± 1.8 pp, low confidence
103Kimi K2.7 Code76.4%estimated ± 1.8 pp, low confidence
104Claude Opus 4.876.3%estimated ± 1.7 pp, low confidence
105Gemini 3.5 Flash76.3%estimated ± 1.7 pp, low confidence
106Qwen3.5 397B76.0%estimated ± 1.8 pp, low confidence
107Nemotron 3.5 Super VL 120B A12B75.8%estimated ± 2.4 pp, low confidence
108Nemotron 3.5 Super VL 120B A12B NVFP475.8%estimated ± 2.4 pp, low confidence
109Solar Pro 475.7%estimated ± 1.8 pp, low confidence
110Muse Spark Open75.5%estimated ± 1.8 pp, low confidence
111A.X K275.4%estimated ± 2.4 pp, low confidence
112GLM-575.2%estimated ± 2.4 pp, low confidence
113Qwen 3.6 Max (preview)75.2%estimated ± 1.8 pp, low confidence
114GPT-5.4 nano74.8%estimated ± 2.4 pp, low confidence
115GLM-5-Turbo74.6%estimated ± 2.4 pp, low confidence
116Kimi K2.573.8%estimated ± 1.8 pp, low confidence
117GPT-5.1-Codex73.7%estimated ± 2.4 pp, low confidence
118GPT-5.1-Codex-Max73.7%estimated ± 2.4 pp, low confidence
119MiniMax M2.773.1%estimated ± 1.8 pp, low confidence
120MiMo-V2-Pro72.5%estimated ± 1.8 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General