GPQA Diamond (AA)

Artificial Analysis' own GPQA Diamond run. Shown for reference; Epoch's run is in the index.

Published by

Artificial Analysis

Category

Reasoning

Index weight

Reference only

Models

525

Data as of

9 Sept 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

1. 1GPT-6 Astra96.3%
2. 2GPT-6 Astra96.1%
3. 3Gemini 3.8 Flash95.3%
4. 4GPT-6 Astra95.0%
5. 5Grok 4.695.0%
6. 6Gemini 3.7 Flash94.5%
7. 7GPT-5.6 Sol94.1%
8. 8Gemini 3.1 Pro94.1%
9. 9Muse Spark 1.394.1%
10. 10GPT-6 Astra93.9%
11. 11Claude Fable 5.193.7%
12. 12Claude Opus 593.7%
13. 13Claude Opus 593.7%
14. 14GPT-5.593.5%
15. 15Kimi K393.5%

SearchAll providersAI21 LabsAI9StarsAlibabaAllen Institute for AIAllenaiAmazonAnt GroupAnt-groupAnthropicApodexArcee AIBaiduByteDanceCelerisChina MobileCohereDeep CogitoDeepSeekGoogleIBMInceptionInclusionAIKimiKorea TelecomKwaiKATLG AI ResearchLiquid AILongCatMBZUAI Institute of Foundation ModelsMetaMicrosoftMiniMaxMistral AIMoonshot AIMotif TechnologiesMultiverse ComputingNanbeigeNaverNex AGINous ResearchNVIDIAOpenAIOpenBMBOpenChat / BerkeleyPerplexityPrime IntellectQwenReka AISapiens AISarvamServiceNowSK TelecomStepFunSwiss AI InitiativeThinking MachinesTII UAETrillion LabsUnknownUpstagexAIXiaomiZhipu AIOpen weights525 of 525

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	GPT-6 AstraOpenAI	96.3%	–	68.6	–
2	GPT-6 AstraOpenAI	96.1%	GPT-6 Astra (max)	69.9	–
3	Gemini 3.8 FlashGoogle	95.3%	Gemini 3.8 Flash (high)	62.4	–
4	GPT-6 AstraOpenAI	95.0%	–	69.5	–
5	Grok 4.6xAI	95.0%	Grok 4.6 (high)	63.6	–
6	Gemini 3.7 FlashGoogle	94.5%	Gemini 3.7 Flash (high)	61.7	–
7	GPT-5.6 SolOpenAI	94.1%	GPT-5.6 Sol (max)	65.3	–
8	Gemini 3.1 ProGoogle	94.1%	–	63.9	–
9	Muse Spark 1.3Meta	94.1%	–	63.9	–
10	GPT-6 AstraOpenAI	93.9%	–	67.4	–
11	Claude Fable 5.1Anthropic	93.7%	Claude Fable 5.1 (max with fallback)	70.0	–
12	Claude Opus 5Anthropic	93.7%	–	69.3	–
13	Claude Opus 5Anthropic	93.7%	–	66.3	–
14	GPT-5.5OpenAI	93.5%	GPT-5.5 (xhigh)	66.6	–
15	Kimi K3Moonshot AIopen	93.5%	Kimi K3 (max)	66.5	–
16	Grok 4.6xAI	93.5%	–	63.8	–
17	Gemini 3.8 FlashGoogle	93.5%	–	63.5	–
18	Grok 4.6xAI	93.5%	–	62.3	–
19	Muse Spark 1.3Meta	93.5%	Muse Spark 1.3 (max)	–	–
20	Qwen3.8 2.4T A95BAlibabaopen	93.5%	–	–	–
21	Claude Fable 5.1Anthropic	93.4%	–	70.3	–
22	Claude Opus 5Anthropic	93.2%	Claude Opus 5 (max)	70.0	–
23	GPT-5.5OpenAI	93.2%	–	66.3	–
24	GPT-5.6 SolOpenAI	93.1%	–	67.5	–
25	GPT-6 AstraOpenAI	93.1%	–	65.8	–
26	Grok 4.5xAI	93.1%	Grok 4.5 (high)	63.4	–
27	MiniMax-M3MiniMaxopen	92.9%	MiniMax-M3	56.9	–
28	GPT-5.6 SolOpenAI	92.8%	–	68.0	–
29	Gemini 3.6 FlashGoogle	92.8%	Gemini 3.6 Flash	60.7	–
30	DeepSeek V4 ProDeepSeekopen	92.8%	DeepSeek V4 Pro 0813 (max)	59.7	–
31	Qwen3 8Alibaba	92.7%	–	63.9	–
32	Claude Fable 5Anthropic	92.6%	Claude Fable 5 (with fallback)	70.4	–
33	GPT-5.6 SolOpenAI	92.6%	–	66.0	–
34	GPT-5.5OpenAI	92.6%	–	64.9	–
35	GPT-5.6 TerraOpenAI	92.5%	GPT-5.6 Terra (max)	62.3	–
36	Qwen3.8-Flash-NextAlibabaopen	92.3%	–	59.3	–
37	Qwen3 7Alibaba	92.3%	–	58.9	–
38	Gemini 3.5 FlashGoogle	92.2%	Gemini 3.5 Flash	60.7	–
39	Gemini 3.7 FlashGoogle	92.1%	–	64.4	–
40	Gemini 3.5 FlashGoogle	92.1%	–	63.9	–
41	Claude Opus 4.8Anthropic	92.0%	Claude Opus 4.8 (max)	64.7	–
42	GPT-5.4OpenAI	92.0%	GPT-5.4 (xhigh)	63.4	–
43	Gemini 3.8 FlashGoogle	92.0%	–	61.9	–
44	Claude Opus 5Anthropic	91.9%	–	63.1	–
45	GLM-5.3Zhipu AIopen	91.7%	GLM-5.3 (max)	–	–
46	GPT-5.3 CodexOpenAI	91.5%	GPT-5.3 Codex (xhigh)	62.4	–
47	Claude Opus 4.7Anthropic	91.4%	Claude Opus 4.7 (max)	65.5	–
48	Deepseek v4 Flash VisionDeepSeek	91.3%	DeepSeek V4 Flash Vision (max)	57.4	–
49	GLM-5.3-FlashZhipu AIopen	91.2%	–	61.0	–
50	Claude Sonnet 5Anthropic	91.1%	Claude Sonnet 5 (max)	61.2	–
51	Kimi K2.6Moonshot AIopen	91.1%	Kimi K2.6	60.6	–
52	Grok 4.20xAI	91.1%	Grok 4.20 0309 v2	57.9	–
53	GPT-5.6 LunaOpenAI	91.1%	GPT-5.6 Luna (max)	56.6	–
54	GPT-5.5OpenAI	91.0%	–	62.3	–
55	GPT-5.6 TerraOpenAI	90.8%	–	64.2	–
56	Gemini 3 ProGoogle	90.8%	Gemini 3 Pro Preview (high)	61.6	–
57	DeepSeek V4 FlashDeepSeekopen	90.8%	DeepSeek V4 Flash 0731 (max)	59.1	–
58	Claude Fable 5.1Anthropic	90.6%	–	69.7	–
59	DeepSeek V4 ProDeepSeekopen	90.5%	high, 0424	60.0	–
60	Qwen 3.8 27BAlibabaopen	90.5%	Qwen3.8 27B (xhigh)	57.1	–

Show 100 more