LiveBench

Average of LiveBench's category scores on the current question set.

Published by

LiveBench

Category

Composite

Index weight

1.0

Models

52

Data as of

9 Sept 2026

LiveBench (livebench.ai), White et al. 2024.

1. 1Claude Fable 5.183.4%
2. 2Claude Fable 583.0%
3. 3GPT-6 Astra82.2%
4. 4Muse Spark 1.381.6%
5. 5GPT-5.6 Sol81.0%
6. 6GPT-5.580.2%
7. 7Claude Opus 580.1%
8. 8Kimi K379.2%
9. 9Gemini 3.7 Flash78.8%
10. 10Qwen3 878.5%
11. 11Grok 4.678.0%
12. 12GPT-5.478.0%
13. 13Muse Spark 1.278.0%
14. 14GPT-5.6 Terra77.9%
15. 15DeepSeek V4 Pro77.4%

SearchAll providersAlibabaAnthropicDeepSeekGoogleMetaMiniMaxMoonshot AINVIDIAOpenAIUnknownxAIZhipu AIOpen weights52 of 52

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	Claude Fable 5.1Anthropic	83.4%	–	70.2	–
2	Claude Fable 5Anthropic	83.0%	–	65.8	–
3	GPT-6 AstraOpenAI	82.2%	–	70.6	–
4	Muse Spark 1.3Meta	81.6%	–	63.9	–
5	GPT-5.6 SolOpenAI	81.0%	–	65.5	–
6	GPT-5.5OpenAI	80.2%	–	63.7	–
7	Claude Opus 5Anthropic	80.1%	–	66.8	–
8	Kimi K3Moonshot AIopen	79.2%	kimi-k3	66.5	–
9	Gemini 3.7 FlashGoogle	78.8%	–	60.8	–
10	Qwen3 8Alibaba	78.5%	–	63.9	–
11	Grok 4.6xAI	78.0%	grok-4.6	63.6	–
12	GPT-5.4OpenAI	78.0%	–	61.3	–
13	Muse Spark 1.2Meta	78.0%	–	61.3	–
14	GPT-5.6 TerraOpenAI	77.9%	–	58.0	–
15	DeepSeek V4 ProDeepSeekopen	77.4%	0813	59.7	–
16	USmaug FlashUnknown	77.4%	–	–	–
17	Gemini 3.1 ProGoogle	77.0%	–	60.7	–
18	USmaug MiniUnknown	76.9%	–	–	–
19	Deepseek v4 Flash VisionDeepSeek	76.8%	–	57.4	–
20	Claude Opus 4.7Anthropic	76.5%	–	57.2	–
21	Claude Opus 4.8Anthropic	76.2%	–	59.4	–
22	Qwen3.8-Flash-NextAlibabaopen	76.2%	–	59.3	–
23	GLM-5.3Zhipu AIopen	76.1%	glm-5.3	–	–
24	Claude Sonnet 5Anthropic	76.0%	–	57.3	–
25	Gemini 3.8 FlashGoogle	75.8%	–	60.1	–
26	Grok 4.5xAI	75.8%	grok-4.5	63.4	–
27	Muse Spark 1.1Meta	75.3%	–	56.4	–
28	Qwen 3.8 27BAlibabaopen	75.3%	–	57.1	–
29	Gemini 3.5 FlashGoogle	74.6%	–	60.4	–
30	GPT-5.2OpenAI	74.6%	–	55.5	–
31	Claude Opus 4.6Anthropic	74.5%	–	61.3	–
32	DeepSeek V4 FlashDeepSeekopen	74.2%	0731	59.1	–
33	GPT-5.2 CodexOpenAI	74.0%	–	60.7	–
34	Gemini 3.6 FlashGoogle	73.6%	–	58.6	–
35	GPT-5.6 LunaOpenAI	73.6%	–	56.5	–
36	GLM-5.2Zhipu AIopen	73.2%	glm-5.2	57.4	–
37	Qwen3 7Alibaba	73.1%	–	58.9	–
38	Claude Sonnet 4.6Anthropic	73.0%	–	51.8	–
39	Claude Opus 4.5Anthropic	72.6%	–	56.6	–
40	GLM-5.3-FlashZhipu AIopen	71.6%	–	61.0	–
41	Kimi K2.6Moonshot AIopen	70.5%	–	–	–
42	GPT-5.4 nanoOpenAI	69.6%	–	–	–
43	UOx AlphaUnknown	69.2%	–	–	–
44	Qwen3.6 PlusAlibabaopen	68.9%	–	59.0	–
45	Kimi K2.7 CodeMoonshot AIopen	68.4%	–	56.2	–
46	Grok Build 0.1xAI	67.8%	–	56.1	–
47	Nemotron 3 Ultra 550B A55BNVIDIAopen	67.4%	–	58.3	–
48	MiniMax-M3MiniMaxopen	67.3%	minimax-m3	56.9	–
49	GPT-5.4 miniOpenAI	66.4%	–	46.0	–
50	Qwen3.6 27BAlibabaopen	64.0%	–	53.2	–
51	Gemini 3.5 Flash LiteGoogle	63.9%	–	44.0	–
52	Grok 4.3xAI	62.3%	grok-4.3	57.8	–