Epoch AI Benchmarking Hub

Shown as Epoch AI publishes it — their metrics, their models. A model name opens its page here; the arrow opens the original board.

Published by

Epoch AI

License

CC-BY-4.0

Updates

continuous

Fetched

2026-10-07

Entries

148 · 10 unmatched

1. GPT 6 Astra 93.7%
2. GPT 6.1 Sol 93.7%
3. Claude Opus 5.5 91.2%
4. Claude Fable 5.1 90.2%
5. GPT 6 Sol 89.8%
6. GPT 5.6 Sol 89.1%
7. Claude Sonnet 5.5 88.8%
8. GPT 5.5 Pro 87.7%
9. Claude Fable 5 87.0%
10. GPT 5.6 Terra 86.0%
11. Claude Opus 5 85.6%
12. GPT 5.5 85.3%

Filter modelsOrder byFrontierMath Tiers 1–3FrontierMath Tier 4SimpleQA VerifiedMatched models only

114 scored · 34 not tested

#	Model	FrontierMath Tiers 1–3 ↑	FrontierMath Tier 4 ↑	SimpleQA Verified ↑
1	GPT 6 AstramaxOpenAI	93.7%	97.6%	75.6%
2	GPT 6.1 SolmaxOpenAI	93.7%	100.0%	73.9%
3	Claude Opus 5.5maxAnthropic	91.2%	95.0%	72.2%
4	Claude Fable 5.1maxAnthropic	90.2%	87.8%	70.8%
5	GPT 6 SolmaxOpenAI	89.8%	90.0%	60.7%
6	GPT 5.6 SolmaxOpenAI	89.1%	82.9%	69.7%
7	Claude Sonnet 5.5maxAnthropic	88.8%	80.5%	46.5%
8	GPT 5.5 ProxhighOpenAI	87.7%	78.0%	—
9	Claude Fable 5maxAnthropic	87.0%	90.2%	—
10	GPT 5.6 TerramaxOpenAI	86.0%	70.7%	43.2%
11	Claude Opus 5maxAnthropic	85.6%	73.2%	59.9%
12	GPT 5.5xhighOpenAI	85.3%	72.5%	63.0%
13	GPT 5.4 Pro2026-03-05 · xhighOpenAI	82.5%	58.5%	46.3%
14	GPT 5.6 LunamaxOpenAI	82.1%	61.0%	41.0%
15	Claude Opus 4.8maxAnthropic	80.0%	56.1%	53.0%
16	GPT 6 LunamaxOpenAI	78.9%	56.1%	41.4%
17	GPT 5.42026-03-05 · xhighOpenAI	78.6%	49.0%	45.1%
18	Qwen 3.8 MaxxhighAlibaba	74.7%	46.3%	45.8%
19	Muse Spark 1.3xhighMeta	74.4%	41.5%	—
20	Muse Spark 1.3maxMeta	74.0%	46.3%	—
21	gpt-5.2-pro-2025-12-11_xhigh2025-12-11 · xhighUnmatched	74.0%	46.0%	—
22	Kimi K3maxKimi	72.2%	39.0%	50.6%
23	Gemini 3.7 FlashhighGoogle	71.6%	36.6%	69.2%
24	Claude Opus 4.7maxAnthropic	70.2%	31.7%	—
25	GLM 5.3maxZ AI	68.8%	29.3%	41.0%
26	Gemini 3.8 FlashhighGoogle	68.4%	22.0%	69.7%
27	GPT 5.22025-12-11 · xhighOpenAI	67.4%	31.7%	37.1%
28	Claude Opus 4.6maxAnthropic	66.0%	26.8%	47.0%
29	Grok 4.6xhighSpaceXAI	66.0%	31.7%	48.9%
30	Claude Sonnet 5maxAnthropic	65.6%	29.3%	33.7%
31	Qwen 3.8 Max0902 · xhighAlibaba	65.6%	34.1%	47.3%
32	DeepSeek V4 Pro0813 · maxDeepSeek	64.6%	26.8%	52.9%
33	Qwen 3.7 MaxAlibaba	64.6%	34.1%	55.8%
34	Gemini 3.5 FlashhighGoogle	62.8%	26.8%	66.2%
35	Gemini 3.1 PropreviewGoogle	59.6%	26.8%	—
36	GLM 5.2maxZ AI	59.2%	29.3%	34.2%
37	Gemini 3.6 FlashhighGoogle	58.9%	22.0%	66.2%
38	DeepSeek V4 Flash0731 · maxDeepSeek	57.5%	24.4%	33.6%
39	Grok 4.5highSpaceXAI	57.2%	24.4%	48.3%
40	Kimi K2.6Kimi	57.2%	25.6%	34.9%
41	GLM 5.3 FlashmaxZ AI	55.8%	17.1%	—
42	gpt-5-pro-2025-10-06_high2025-10-06 · highUnmatched	55.8%	19.5%	—
43	GPT 52025-08-07 · highOpenAI	55.4%	22.0%	50.1%
44	GLM 5.2lowZ AI	54.7%	—	—
45	Kimi K2.7 CodeKimi	54.0%	12.2%	36.5%
46	Grok 4.7xhighSpaceXAI	53.0%	17.1%	56.0%
47	Gemini 3 FlashpreviewGoogle	51.2%	17.1%	—
48	GPT 5.4 mini2026-03-17 · xhighOpenAI	51.2%	9.8%	—
49	GPT 5 mini2025-08-07 · highOpenAI	46.7%	12.2%	21.6%
50	Inkling SmallxhighThinking Machines	46.3%	17.1%	19.1%
51	DeepSeek V4 PromaxDeepSeek	45.3%	2.4%	47.0%
52	GPT 5.4 nano2026-03-17 · highOpenAI	44.9%	12.2%	11.7%
53	Grok 4.200309 · onSpaceXAI	44.9%	17.1%	30.2%
54	Grok 4.3highSpaceXAI	42.8%	14.6%	33.2%
55	GLM 5.2offZ AI	42.5%	—	—
56	GPT 5.6 LunalowOpenAI	41.4%	—	—
57	GPT 5.6 LunaoffOpenAI	39.6%	—	—
58	Qwen 3.6 PlusAlibaba	38.2%	—	44.1%
59	GPT 52025-08-07 · lowOpenAI	37.2%	—	—
60	GLM 5.1Z AI	36.8%	—	34.0%
61	o4 mini2025-04-16 · highOpenAI	36.1%	4.9%	18.8%
62	Qwen 3.627BAlibaba	35.1%	—	—
63	Claude Opus 4.520251101 · on · 32000Anthropic	34.4%	4.9%	45.7%
64	Qwen 3.7 PlusoffAlibaba	34.4%	—	—
65	Qwen 3.627B · offAlibaba	34.0%	—	—
66	InklingxhighThinking Machines	33.3%	4.9%	40.3%
67	o32025-04-16 · highOpenAI	33.3%	—	49.4%
68	Qwen 3.6 PlusoffAlibaba	32.3%	—	—
69	Qwen 3.5397B A17B · offAlibaba	31.2%	—	—
70	o32025-04-16 · mediumOpenAI	29.8%	—	—
71	Qwen 3.5397B A17BAlibaba	29.5%	—	—
72	o4 mini2025-04-16 · mediumOpenAI	28.8%	—	—
73	Gemini 3.1 Flash LitehighGoogle	27.7%	—	—
74	GPT 5.5 InstantOpenAI	26.3%	2.4%	—
75	Gemini 3.5 Flash LitehighGoogle	26.0%	0.0%	—
76	GLM 5.1offZ AI	24.9%	—	—
77	Gemini 2.5 ProGoogle	24.6%	0.0%	—
78	GPT 5.4 mini2026-03-17 · lowOpenAI	24.6%	—	—
79	Claude Sonnet 4.520250929 · on · 32000Anthropic	23.9%	2.4%	—
80	Gemini 3.1 Flash LitelowGoogle	22.5%	—	—
81	qwen3.6-flash_noneoffUnmatched	22.5%	—	—
82	Gemini 3.1 Flash LiteminimalGoogle	21.4%	—	—
83	GPT 5.4 nano2026-03-17 · lowOpenAI	20.4%	—	—
84	Qwen 3.635B A3B · offAlibaba	20.4%	—	—
85	GPT 5 nano2025-08-07 · highOpenAI	20.0%	2.4%	11.7%
86	o32025-04-16 · lowOpenAI	19.3%	—	—
87	qwen3.7-flashUnmatched	19.3%	—	—
88	qwen3.7-flash_noneoffUnmatched	19.3%	—	—
89	Qwen 3 Max2025-09-23Alibaba	18.9%	—	48.7%
90	o3 mini2025-01-31 · highOpenAI	18.6%	0.0%	15.3%
91	GPT 52025-08-07 · minimalOpenAI	18.2%	—	—
92	GPT 5 mini2025-08-07 · lowOpenAI	18.2%	—	—
93	Qwen 3.5 FlashoffAlibaba	18.2%	—	—
94	Qwen 3.635B A3BAlibaba	17.5%	—	—
95	GPT 5.4 mini2026-03-17 · offOpenAI	17.2%	—	—
96	qwen3.6-flashUnmatched	17.2%	—	15.9%
97	o4 mini2025-04-16 · lowOpenAI	16.1%	—	19.6%
98	o12024-12-17 · highOpenAI	14.7%	—	41.1%
99	Claude Opus 4.120250805 · on · 32000Anthropic	12.6%	2.4%	—
100	o3 mini2025-01-31 · mediumOpenAI	10.5%	—	—

Show all 148 rows

What each column measures

FrontierMath Tiers 1–3

Higher is better · Reasoning and knowledge

FrontierMath Tier 4

Higher is better · Reasoning and knowledge

SimpleQA Verified

Higher is better · Factuality