Coding

AA LiveCodeBench leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

As of 2026-10-29, the highest measured score on AA LiveCodeBench is 91.7% by Gemini 3 Pro. 76 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Gemini 3.5 Flash	92.8%	estimated ± 3.4 pp, low confidence
2	GPT-5.4 mini	92.7%	estimated ± 3.4 pp, low confidence
3	Gemini 3 Pro	91.7%	measured
4	GPT-5.2-Codex	90.6%	estimated ± 3.4 pp, low confidence
5	Kimi K2.6	90.6%	estimated ± 3.4 pp, low confidence
6	GLM-4.7	89.4%	measured
7	Gemini 3.1 Pro	89.4%	estimated ± 3.4 pp, low confidence
8	GLM-5.1	89.2%	estimated ± 3.4 pp, low confidence
9	MiniMax M2.7	88.3%	estimated ± 3.4 pp, low confidence
10	GPT-5.4 nano	88.1%	estimated ± 3.4 pp, low confidence
11	Apodex 1.1	88.1%	estimated ± 4.7 pp, low confidence
12	Apodex 1.1 Mini	88.1%	estimated ± 4.7 pp, low confidence
13	Hy3	88.1%	estimated ± 4.7 pp, low confidence
14	Hy3 Preview	88.1%	estimated ± 4.7 pp, low confidence
15	Kimi K2.7 Code	88.1%	estimated ± 4.7 pp, low confidence
16	MiMo-V2.5-Pro	88.1%	estimated ± 4.7 pp, low confidence
17	MiniMax M3	88.1%	estimated ± 4.7 pp, low confidence
18	Quasar 438B	88.1%	estimated ± 4.7 pp, low confidence
19	Qwen3.7 Max	88.1%	estimated ± 4.7 pp, low confidence
20	Qwen3.7 Plus	88.1%	estimated ± 4.7 pp, low confidence
21	Qwen3.8-27B	88.1%	estimated ± 4.7 pp, low confidence
22	Qwen3.6-27B	88.1%	estimated ± 4.7 pp, low confidence
23	Inkling-Small	88.1%	estimated ± 4.7 pp, low confidence
24	Inkling	88.1%	estimated ± 4.7 pp, low confidence
25	Ling 3.0 Flash	88.1%	estimated ± 4.7 pp, low confidence
26	MiMo-V2-Flash	88.1%	estimated ± 4.7 pp, low confidence
27	Gemini 3.5 Flash-Lite	88.1%	estimated ± 4.7 pp, low confidence
28	Nemotron 3 Ultra	88.1%	estimated ± 4.7 pp, low confidence
29	Muse Glimmer 30B	88.1%	estimated ± 4.7 pp, low confidence
30	Mistral Medium 3.5 128B	88.1%	estimated ± 4.7 pp, low confidence
31	Kimi K2.5	88.1%	estimated ± 4.7 pp, low confidence
32	Qwen3.5-122B-A10B	88.1%	estimated ± 4.7 pp, low confidence
33	Gemma 4 31B	88.1%	estimated ± 4.7 pp, low confidence
34	Grok 4.3	88.1%	estimated ± 4.7 pp, low confidence
35	Qwen3.6-35B-A3B	88.1%	estimated ± 4.7 pp, low confidence
36	Qwen3.6 Plus	88.0%	estimated ± 3.4 pp, low confidence
37	GPT-OSS 120B	87.8%	measured
38	o1	87.7%	estimated ± 4.7 pp, low confidence
39	Step 3.7 Flash	87.6%	estimated ± 4.7 pp, low confidence
40	GLM-5 (Reasoning)	87.5%	estimated ± 3.4 pp, low confidence
41	Claude Sonnet 4.5 Thinking	87.4%	estimated ± 3.4 pp, low confidence
42	GPT-5.1-Codex-Max	87.3%	estimated ± 3.4 pp, low confidence
43	Gemma 4 26B A4B	87.3%	estimated ± 4.7 pp, low confidence
44	Claude Opus 4.5 Thinking	87.1%	measured
45	Gemini 3 Flash	86.9%	estimated ± 3.4 pp, low confidence
46	GPT-5.1	86.8%	measured
47	Muse Spark	86.8%	estimated ± 3.4 pp, low confidence
48	Kimi K2.5 (Reasoning)	86.3%	estimated ± 3.4 pp, low confidence
49	Qwen3.5 Plus	86.0%	estimated ± 3.4 pp, low confidence
50	o4-mini	85.9%	measured
51	MiniMax M2.5	85.8%	estimated ± 3.4 pp, low confidence
52	GPT-5 mini	85.6%	estimated ± 3.4 pp, low confidence
53	GPT-5.1-Codex	85.4%	estimated ± 3.4 pp, low confidence
54	Claude Haiku 4.5 Thinking	85.1%	estimated ± 3.4 pp, low confidence
55	GPT-5 (high)	84.6%	measured
56	o1-preview	84.0%	estimated ± 4.7 pp, low confidence
57	K-Exaone	84.0%	estimated ± 4.7 pp, low confidence
58	Gemma 4 12B	84.0%	estimated ± 4.7 pp, low confidence
59	Command A+	84.0%	estimated ± 4.7 pp, low confidence
60	Nemotron 3.5 Lightning 30B A3B NVFP4	84.0%	estimated ± 4.7 pp, low confidence
61	Mistral Small 4	84.0%	estimated ± 4.7 pp, low confidence
62	Mistral Small 4 (Reasoning)	84.0%	estimated ± 4.7 pp, low confidence
63	Trinity-Large-Preview	84.0%	estimated ± 4.7 pp, low confidence
64	Trinity-Large-Thinking	84.0%	estimated ± 4.7 pp, low confidence
65	Ling 2.6 Flash	84.0%	estimated ± 4.7 pp, low confidence
66	Gemini 1.5 Pro	84.0%	estimated ± 4.7 pp, low confidence
67	DeepSeek V3	84.0%	estimated ± 4.7 pp, low confidence
68	Granite 4.2 8B	84.0%	estimated ± 4.7 pp, low confidence
69	Celeris-1	84.0%	estimated ± 4.7 pp, low confidence
70	Claude 3 Opus	84.0%	estimated ± 4.7 pp, low confidence
71	GPT-4.1 mini	84.0%	estimated ± 4.7 pp, low confidence
72	GPT-4 Turbo	84.0%	estimated ± 4.7 pp, low confidence
73	GPT-OSS 20B	84.0%	estimated ± 4.7 pp, low confidence
74	Llama 4 Maverick	84.0%	estimated ± 4.7 pp, low confidence
75	Mistral Large 3	84.0%	estimated ± 4.7 pp, low confidence
76	Nemotron 3 Nano 30B	84.0%	estimated ± 4.7 pp, low confidence
77	Nemotron 3 Nano Omni 30B A3B	84.0%	estimated ± 4.7 pp, low confidence
78	Ultravox v0.6 Llama 3.3 70B	84.0%	estimated ± 4.7 pp, low confidence
79	DeepSeek V3.2 (Thinking)	83.7%	estimated ± 3.4 pp, low confidence
80	Grok 4.20	83.5%	estimated ± 3.4 pp, low confidence
81	Qwen3 Max	83.4%	estimated ± 3.4 pp, low confidence
82	GLM-4.6	83.3%	estimated ± 3.4 pp, low confidence
83	Grok 4 Fast (Reasoning)	83.2%	measured
84	Gemini 3.1 Flash-Lite	82.7%	estimated ± 3.4 pp, low confidence
85	Grok 4.1 Fast (Reasoning)	82.2%	measured
86	Grok 4	81.9%	measured
87	o3	80.8%	measured
88	Gemini 2.5 Pro	80.1%	measured

AA LiveCodeBench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General