benchgap
Agentic · tools

Agents' Last Exam leaderboard

As of 2026-11-21, the highest measured score on Agents' Last Exam is 60.0% by GPT-6.1 Terra. 92 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.1 Terra60.0%measured
2GPT-6 Astra59.3%measured
3GPT-6 Sol56.4%measured
4GPT-6.2 Sol55.0%estimated ± 11.0 pp, low confidence
5Muse Wonder Full53.9%estimated ± 11.0 pp, low confidence
6Qwen3.8 Max52.4%measured
7Claude Fable 5.252.3%estimated ± 10.7 pp, low confidence
8Gemini 4 Neon52.3%estimated ± 10.7 pp, low confidence
9Gemini 4 Carbon51.7%estimated ± 9.6 pp, low confidence
10GPT-6 Cyber51.7%estimated ± 9.6 pp, low confidence
11Claude Mythos 5.251.7%estimated ± 9.6 pp, medium confidence
12GPT-6.1 Sol51.7%estimated ± 9.6 pp, medium confidence
13Claude Opus 5.551.7%estimated ± 9.6 pp, medium confidence
14GPT-5.6 Cyber51.7%estimated ± 9.6 pp, medium confidence
15Claude Mythos 5.151.7%estimated ± 9.6 pp, medium confidence
16Qwen3.8-Flash-Next51.2%measured
17Gemini 4 Hydrogen50.1%estimated ± 10.7 pp, low confidence
18GPT-6.2 Astra49.6%estimated ± 10.7 pp, low confidence
19Claude Sonnet 5.549.1%estimated ± 10.7 pp, low confidence
20Claude Opus 548.4%estimated ± 11.0 pp, low confidence
21Claude Fable 547.4%estimated ± 11.0 pp, low confidence
22Qwen4 Max47.0%estimated ± 12.9 pp, low confidence
23Claude Fable 5.146.3%estimated ± 11.0 pp, low confidence
24Muse Wonder45.8%estimated ± 10.7 pp, low confidence
25Kimi K3.1 Pro43.2%measured
26Qwen3.8-27B42.9%measured
27Hy441.0%measured
28Claude Haiku 5.540.9%estimated ± 10.7 pp, low confidence
29Inkling 2 Small40.8%estimated ± 11.6 pp, low confidence
30GPT-5.5 Pro40.7%estimated ± 11.6 pp, low confidence
31Grok 4.740.1%estimated ± 10.7 pp, low confidence
32GPT-5.4 Pro39.5%estimated ± 11.6 pp, low confidence
33Gemini 4 Argon39.5%measured
34LongCat-2.538.5%measured
35ERNIE 5.5 Flash Preview36.8%estimated ± 11.6 pp, low confidence
36Gemini 3.1 Pro36.5%estimated ± 11.0 pp, low confidence
37Grok 4.636.5%estimated ± 11.0 pp, low confidence
38Kimi K336.5%estimated ± 11.0 pp, low confidence
39Gemini 3.8 Flash36.2%estimated ± 10.7 pp, low confidence
40Mistral Large 436.0%estimated ± 10.7 pp, low confidence
41Ornith-1.5-397B35.7%estimated ± 11.6 pp, low confidence
42DeepSeek V4.1 Flash 101735.4%measured
43Claude Haiku 4.535.4%estimated ± 11.0 pp, low confidence
44Claude Opus 4.835.4%estimated ± 11.0 pp, low confidence
45Claude Sonnet 535.4%estimated ± 11.0 pp, low confidence
46Gemini 3.5 Flash35.4%estimated ± 11.0 pp, low confidence
47Gemini 3.6 Flash35.4%estimated ± 11.0 pp, low confidence
48GLM-5.235.4%estimated ± 11.0 pp, low confidence
49GPT-5.3-Codex-Spark35.4%estimated ± 11.0 pp, low confidence
50Grok 4.2035.4%estimated ± 11.0 pp, low confidence
51Llama 4 Maverick35.4%estimated ± 11.0 pp, low confidence
52MiMo-V2.5-Pro35.4%estimated ± 11.0 pp, low confidence
53MiniMax M335.4%estimated ± 11.0 pp, low confidence
54Muse Ember35.4%estimated ± 11.0 pp, low confidence
55Muse Spark 1.235.4%estimated ± 11.0 pp, low confidence
56Muse Spark 1.335.4%estimated ± 11.0 pp, low confidence
57Muse Spark Open35.4%estimated ± 11.0 pp, low confidence
58Nemotron 3 Ultra35.4%estimated ± 11.0 pp, low confidence
59MiMo-V2.6-Pro31.6%measured
60dots3-note Preview31.5%estimated ± 11.6 pp, low confidence
61Kimi K2.631.3%estimated ± 11.6 pp, low confidence
62GPT-6 Luna30.7%estimated ± 9.6 pp, medium confidence
63Beam 1.0 501B-A23B30.5%estimated ± 11.4 pp, low confidence
64Atria Dawn Preview30.1%estimated ± 3.1 pp, high confidence
65Ling 3.1 Flash29.9%estimated ± 3.1 pp, high confidence
66MiniMax M3.1 Flash29.6%measured
67Sage-A1.529.6%estimated ± 11.6 pp, low confidence
68Step 5 Preview29.5%measured
69Fugu Cyber28.9%estimated ± 3.3 pp, medium confidence
70Gemini 3.8 Flash Cyber28.7%estimated ± 3.3 pp, medium confidence
71GLM-5.328.5%measured
72GPT-5.6 Sol28.3%estimated ± 3.3 pp, medium confidence
73Claude Mythos 528.1%estimated ± 3.3 pp, medium confidence
74Gemini 3.5 Flash Cyber27.9%estimated ± 3.3 pp, medium confidence
75Claude Mythos Preview27.9%estimated ± 3.3 pp, medium confidence
76Ling 3.0 Flash Sante27.8%estimated ± 11.6 pp, low confidence
77MiMo-V2.6-Flash27.6%measured
78GPT-5.527.5%estimated ± 3.3 pp, medium confidence
79GPT-5.6 Terra27.5%estimated ± 3.3 pp, medium confidence
80GPT-5.426.7%estimated ± 3.3 pp, medium confidence
81Inkling 2 Mini26.6%estimated ± 11.6 pp, low confidence
82GPT-5.6 Luna26.3%estimated ± 3.3 pp, medium confidence
83Gemini 3.7 Flash26.3%measured
84GLM-5.3-Flash26.3%measured
85DeepSeek V4 Pro25.7%measured
86Nemotron 3.5 Super VL 120B A12B25.5%estimated ± 11.6 pp, low confidence
87DeepSeek V4 Flash 073125.2%measured
88Nemotron 3.5 Super VL 120B A12B NVFP425.0%estimated ± 11.6 pp, low confidence
89Claude Opus 4.724.9%estimated ± 3.3 pp, low confidence
90Inkling-Small24.9%estimated ± 11.6 pp, low confidence
91Inkling24.6%estimated ± 11.6 pp, low confidence
92Beam 1.0 501B-A23B FP824.2%estimated ± 11.6 pp, low confidence
93GLM-5.123.6%estimated ± 3.3 pp, low confidence
94Step 3.7 Flash23.4%estimated ± 11.6 pp, low confidence
95Beam 1.0 501B-A23B NVFP423.2%estimated ± 11.6 pp, low confidence
96Agents-A123.1%estimated ± 11.6 pp, low confidence
97Claude Opus 4.623.0%estimated ± 3.3 pp, low confidence
98Hy4 preview22.8%measured
99Claude Sonnet 4.622.5%estimated ± 3.3 pp, low confidence
100Muse Spark 1.120.6%estimated ± 3.3 pp, low confidence
101Ling 3.0 Flash20.4%estimated ± 11.6 pp, low confidence
102Nemotron 3.5 Bolt 74B-A9B BF1618.8%estimated ± 11.6 pp, low confidence
103Qwen3.5 397B18.1%estimated ± 11.6 pp, low confidence
104Ornith-1.5-35B-A3B17.3%estimated ± 11.6 pp, low confidence
105Agents-A1-4B16.8%estimated ± 11.6 pp, low confidence
106Nemotron 3.5 Bolt 74B-A9B16.8%estimated ± 11.6 pp, low confidence
107GPT-5.216.3%estimated ± 11.6 pp, low confidence
108Qwen3.5-122B-A10B15.3%estimated ± 11.6 pp, low confidence
109GLM-514.6%estimated ± 11.6 pp, low confidence
110Qwen3.5-27B14.2%estimated ± 11.6 pp, low confidence
111Qwen3.5-35B-A3B14.2%estimated ± 11.6 pp, low confidence
112Kimi K2.514.0%estimated ± 11.6 pp, low confidence
113Ornith-1.5-9B12.7%estimated ± 11.6 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General