Best frontier models right now

One list, in order. Each model's score is its average placing on Arena's overall, hard-problem, coding and agent boards (100 = #1 everywhere). Tap a model for its reference page.

#	Model	Score	Overall	Hard	Coding	Agents
1	👑 Claude Opus 5.5Anthropic	97	#2	#2	#1	#1
2	GPT-6 AstraOpenAI	95	—	—	#2	#2
3	Claude Fable 5.1Anthropic	84	#3	#5	#5	#3
4	Gemini 4 ArgonGoogle DeepMind	83	#1	#1	#8	#7
5	GPT-6.1 SolOpenAI	82	—	—	#4	#5
6	Claude Opus 5Anthropic	75	#4	#4	#6	#9
7	GPT-6 SolOpenAI	71	—	—	#7	#6
8	Claude Fable 5Anthropic	50	#8	#8	#18	#8
9	Claude Sonnet 5.5Anthropic	47	#24	#13	#3	#4
10	Muse Spark 1.3Meta AI	46	#10	#10	#11	#14
11	XMiMo V2.6 ProXiaomi · open weights	42	#7	#6	#17	#18
12	Claude Opus 4.6Anthropic	40	#5	#3	#36	—
13	AQwen 3.8Alibaba	33	#12	#11	#10	#22
14	MAKimi K3Moonshot AI · open weights	26	#20	#12	#12	#16
15	Gemini 3.8 FlashGoogle DeepMind	24	#6	#7	#28	#21
16	Claude Opus 4.7Anthropic	22	#9	#9	#33	—
17	Gemini 3.5 FlashGoogle DeepMind	21	#14	#18	—	—
18	Grok 4.7xAI	18	—	—	#13	#20
19	THy4Tencent · open weights	18	—	—	#16	#17
20	GPT-5.5OpenAI	14	#22	#17	—	#13
21	Gemini 3.1 ProGoogle DeepMind	8	#15	#22	—	—
22	Gemini 3.7 FlashGoogle DeepMind	-1	#11	#14	#26	#36
23	ZGLM 5.3Z.ai · open weights	-1	#23	#16	#19	#23
24	Claude Sonnet 5Anthropic	-4	—	—	#39	#12
25	Muse Spark 1.2Meta AI	-4	#13	#20	—	#37

Source: Arena (arena.ai) public leaderboards, published Oct 9, 2026. A model must be on at least two of the four boards. “—” means it isn't on that board, so it's judged on the rest.