Best AI for coding: all 77 ranked models

Ordered by coding votes. Score shown: how often people preferred its answers to coding questions. Higher is better, and neighbouring models are often too close to separate: check the range before reading much into a one-place gap.

This is one test, not a recommendation. Our pick for code something also looks at web-building votes, fixing real software bugs and difficult coding tasks, and flags close calls. See the options worth considering.

#	Model	Position	Score	Rating	Source
1	Claude Opus 5.5 Best-supported choiceAnthropic	Ranked 1st of 77	1,54795% range 1,535 to 1,559	Top tier	Source for Claude Opus 5.5 code something score (lmarena)
2	Gemini 4 Argon Good-enough alternativeGoogle DeepMind	Ranked 2nd of 77	1,53895% range 1,527 to 1,550	Top tier	Source for Gemini 4 Argon code something score (lmarena)
3	MiMo-V2.6-Pro Good-enough alternative Open weightsXiaomi	Ranked 3rd of 77	1,53495% range 1,516 to 1,551	Top tier	Source for MiMo-V2.6-Pro code something score (lmarena)
4	Claude Opus 5Anthropic	Ranked 4th of 77	1,53295% range 1,525 to 1,539	Top tier	Source for Claude Opus 5 code something score (lmarena)
5	Claude Opus 4.6Anthropic	Ranked 5th of 77	1,53195% range 1,526 to 1,537	Top tier	Source for Claude Opus 4.6 code something score (lmarena)
6	Claude Fable 5.1Anthropic	Ranked 6th of 77	1,53095% range 1,519 to 1,541	Top tier	Source for Claude Fable 5.1 code something score (lmarena)
7	Claude Opus 4.7Anthropic	Ranked 7th of 77	1,51995% range 1,513 to 1,525	Top tier	Source for Claude Opus 4.7 code something score (lmarena)
8	Muse WonderMeta	Ranked 8th of 77	1,51895% range 1,505 to 1,532	Top tier	Source for Muse Wonder code something score (lmarena)
9	Claude Fable 5Anthropic	Ranked 9th of 77	1,51895% range 1,511 to 1,524	Top tier	Source for Claude Fable 5 code something score (lmarena)
10	Claude Sonnet 5.5Anthropic	Ranked 10th of 77	1,51295% range 1,499 to 1,525	Top tier	Source for Claude Sonnet 5.5 code something score (lmarena)
11	Muse Spark 1.3Meta	Ranked 11th of 77	1,50995% range 1,501 to 1,518	Top tier	Source for Muse Spark 1.3 code something score (lmarena)
12	Gemini 3.8 FlashGoogle DeepMind	Ranked 12th of 77	1,50995% range 1,503 to 1,516	Top tier	Source for Gemini 3.8 Flash code something score (lmarena)
13	GLM-5.3-Flash Open weightsZ.ai (Zhipu AI)	Ranked 13th of 77	1,50995% range 1,502 to 1,516	Top tier	Source for GLM-5.3-Flash code something score (lmarena)
14	Kimi K3 Open weightsMoonshot	Ranked 14th of 77	1,50895% range 1,501 to 1,515	Top tier	Source for Kimi K3 code something score (lmarena)
15	Muse Spark 1.2Meta	Ranked 15th of 77	1,50695% range 1,495 to 1,517	Top tier	Source for Muse Spark 1.2 code something score (lmarena)

Show the other 62 ranked models

#	Model	Position	Score	Rating	Source
16	Qwen3.8-MaxAlibaba	Ranked 16th of 77	1,50495% range 1,496 to 1,511	Top tier	Source for Qwen3.8-Max code something score (lmarena)
17	Claude Sonnet 4.6Anthropic	Ranked 17th of 77	1,50395% range 1,498 to 1,509	Top tier	Source for Claude Sonnet 4.6 code something score (lmarena)
18	DeepSeek-V4.1-Flash-20260910 Open weightsDeepSeek	Ranked 18th of 77	1,50195% range 1,491 to 1,511	Top tier	Source for DeepSeek-V4.1-Flash-20260910 code something score (lmarena)
19	MiMo-V2.5-Pro Open weightsXiaomi	Ranked 19th of 77	1,50195% range 1,496 to 1,507	Top tier	Source for MiMo-V2.5-Pro code something score (lmarena)
20	Gemini 3.7 FlashGoogle DeepMind	Ranked 20th of 77	1,50095% range 1,493 to 1,507	Top tier	Source for Gemini 3.7 Flash code something score (lmarena)
21	Claude Opus 4.5 (20251101)Anthropic	Ranked 21st of 77	1,49895% range 1,493 to 1,504	Strong	Source for Claude Opus 4.5 (20251101) code something score (lmarena)
22	MiMo-V2.6-Flash Open weightsXiaomi	Ranked 22nd of 77	1,49895% range 1,488 to 1,508	Strong	Source for MiMo-V2.6-Flash code something score (lmarena)
23	Muse Spark 1.1Meta	Ranked 23rd of 77	1,49895% range 1,491 to 1,504	Strong	Source for Muse Spark 1.1 code something score (lmarena)
24	GLM-5.3 Open weightsZ.ai (Zhipu AI)	Ranked 24th of 77	1,49795% range 1,489 to 1,505	Strong	Source for GLM-5.3 code something score (lmarena)
25	Qwen3.7-Max-PreviewAlibaba	Ranked 25th of 77	1,49795% range 1,479 to 1,515	Strong	Source for Qwen3.7-Max-Preview code something score (lmarena)
26	GPT-5.4OpenAI	Ranked 26th of 77	1,49695% range 1,491 to 1,502	Strong	Source for GPT-5.4 code something score (lmarena)
27	GPT-5.6 SolOpenAI	Ranked 27th of 77	1,49595% range 1,489 to 1,502	Strong	Source for GPT-5.6 Sol code something score (lmarena)
28	GPT-5.5OpenAI	Ranked 28th of 77	1,49495% range 1,489 to 1,500	Strong	Source for GPT-5.5 code something score (lmarena)
29	Gemini 3.5 FlashGoogle DeepMind	Ranked 29th of 77	1,49395% range 1,487 to 1,499	Strong	Source for Gemini 3.5 Flash code something score (lmarena)
30	Gemini 3.6 FlashGoogle DeepMind	Ranked 30th of 77	1,49195% range 1,485 to 1,498	Strong	Source for Gemini 3.6 Flash code something score (lmarena)
31	Claude Opus 4.8Anthropic	Ranked 31st of 77	1,49195% range 1,485 to 1,497	Strong	Source for Claude Opus 4.8 code something score (lmarena)
32	Kimi K2.6 Open weightsMoonshot	Ranked 32nd of 77	1,48995% range 1,482 to 1,496	Strong	Source for Kimi K2.6 code something score (lmarena)
33	GPT-6 AstraOpenAI	Ranked 33rd of 77	1,48895% range 1,477 to 1,499	Strong	Source for GPT-6 Astra code something score (lmarena)
34	ERNIE 5.1Baidu	Ranked 34th of 77	1,48895% range 1,481 to 1,495	Strong	Source for ERNIE 5.1 code something score (lmarena)
35	GLM-5.2 Open weightsZ.ai (Zhipu AI)	Ranked 35th of 77	1,48795% range 1,481 to 1,493	Strong	Source for GLM-5.2 code something score (lmarena)
36	Qwen3.5-Max-PreviewAlibaba	Ranked 36th of 77	1,48795% range 1,479 to 1,495	Strong	Source for Qwen3.5-Max-Preview code something score (lmarena)
37	GPT-6.1 SolOpenAI	Ranked 37th of 77	1,48695% range 1,472 to 1,500	Strong	Source for GPT-6.1 Sol code something score (lmarena)
38	GLM-5.1 Open weightsZ.ai (Zhipu AI)	Ranked 38th of 77	1,48695% range 1,481 to 1,491	Strong	Source for GLM-5.1 code something score (lmarena)
39	GPT-5.6 TerraOpenAI	Ranked 39th of 77	1,48595% range 1,479 to 1,492	Strong	Source for GPT-5.6 Terra code something score (lmarena)
40	Claude Sonnet 5Anthropic	Ranked 40th of 77	1,48495% range 1,478 to 1,490	Middle	Source for Claude Sonnet 5 code something score (lmarena)
41	Gemini 3.1 Pro PreviewGoogle DeepMind	Ranked 41st of 77	1,48495% range 1,479 to 1,488	Middle	Source for Gemini 3.1 Pro Preview code something score (lmarena)
42	amazon-nova-experimental-chat (26-02-10)Amazon	Ranked 42nd of 77	1,48395% range 1,464 to 1,502	Middle	Source for amazon-nova-experimental-chat (26-02-10) code something score (lmarena)
43	Muse SparkMeta	Ranked 43rd of 77	1,48195% range 1,471 to 1,491	Middle	Source for Muse Spark code something score (lmarena)
44	Gemini 3 ProGoogle DeepMind	Ranked 44th of 77	1,48195% range 1,474 to 1,488	Middle	Source for Gemini 3 Pro code something score (lmarena)
45	Mistral Large 4Mistral AI	Ranked 45th of 77	1,47895% range 1,467 to 1,490	Middle	Source for Mistral Large 4 code something score (lmarena)
46	Grok 4.5xAI	Ranked 46th of 77	1,47495% range 1,467 to 1,481	Middle	Source for Grok 4.5 code something score (lmarena)
47	Kimi K2.5 Open weightsMoonshot	Ranked 47th of 77	1,47495% range 1,469 to 1,479	Middle	Source for Kimi K2.5 code something score (lmarena)
48	Qwen3.7-PlusAlibaba	Ranked 48th of 77	1,47395% range 1,466 to 1,479	Middle	Source for Qwen3.7-Plus code something score (lmarena)
49	Dola Seed 2.0 ProByteDance	Ranked 49th of 77	1,47295% range 1,467 to 1,477	Middle	Source for Dola Seed 2.0 Pro code something score (lmarena)
50	Qwen3.6-Max-PreviewAlibaba	Ranked 50th of 77	1,47195% range 1,456 to 1,486	Middle	Source for Qwen3.6-Max-Preview code something score (lmarena)
51	DeepSeek-V4-Pro Open weightsDeepSeek	Ranked 51st of 77	1,46995% range 1,463 to 1,475	Middle	Source for DeepSeek-V4-Pro code something score (lmarena)
52	DeepSeek-V4-Pro (High, 20260813) Open weightsDeepSeek	Ranked 52nd of 77	1,46995% range 1,457 to 1,480	Middle	Source for DeepSeek-V4-Pro (High, 20260813) code something score (lmarena)
53	NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) Open weightsNVIDIA	Ranked 53rd of 77	1,46895% range 1,457 to 1,479	Middle	Source for NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) code something score (lmarena)
54	Qwen3.6-PlusAlibaba	Ranked 54th of 77	1,46795% range 1,461 to 1,473	Middle	Source for Qwen3.6-Plus code something score (lmarena)
55	GPT-5.6 LunaOpenAI	Ranked 55th of 77	1,46695% range 1,460 to 1,473	Middle	Source for GPT-5.6 Luna code something score (lmarena)
56	Grok 4.6xAI	Ranked 56th of 77	1,46495% range 1,457 to 1,472	Middle	Source for Grok 4.6 code something score (lmarena)
57	Qwen3.5-397B-A17B Open weightsAlibaba	Ranked 57th of 77	1,46495% range 1,459 to 1,469	Middle	Source for Qwen3.5-397B-A17B code something score (lmarena)
58	GLM-5 Open weightsZ.ai (Zhipu AI)	Ranked 58th of 77	1,46195% range 1,454 to 1,468	Middle	Source for GLM-5 code something score (lmarena)
59	Gemini 3 FlashGoogle DeepMind	Ranked 59th of 77	1,45995% range 1,452 to 1,467	Behind the leaders	Source for Gemini 3 Flash code something score (lmarena)
60	Gemma 4 31B Open weightsGoogle DeepMind	Ranked 60th of 77	1,45995% range 1,444 to 1,474	Behind the leaders	Source for Gemma 4 31B code something score (lmarena)
61	Grok 4.20 Beta (Reasoning, 0309)xAI	Ranked 61st of 77	1,45995% range 1,453 to 1,464	Behind the leaders	Source for Grok 4.20 Beta (Reasoning, 0309) code something score (lmarena)
62	Grok 4.20 Multi-Agent Beta (0309)xAI	Ranked 62nd of 77	1,45795% range 1,452 to 1,463	Behind the leaders	Source for Grok 4.20 Multi-Agent Beta (0309) code something score (lmarena)
63	ERNIE 5.0 (0110)Baidu	Ranked 63rd of 77	1,45595% range 1,448 to 1,461	Behind the leaders	Source for ERNIE 5.0 (0110) code something score (lmarena)
64	Claude Haiku 5.5Anthropic	Ranked 64th of 77	1,45495% range 1,440 to 1,469	Behind the leaders	Source for Claude Haiku 5.5 code something score (lmarena)
65	GLM-4.7 Open weightsZ.ai (Zhipu AI)	Ranked 65th of 77	1,45495% range 1,442 to 1,466	Behind the leaders	Source for GLM-4.7 code something score (lmarena)
66	GPT-5.1OpenAI	Ranked 66th of 77	1,45395% range 1,446 to 1,460	Behind the leaders	Source for GPT-5.1 code something score (lmarena)
67	DeepSeek-V4-Pro Preview Open weightsDeepSeek	Ranked 67th of 77	1,45395% range 1,447 to 1,459	Behind the leaders	Source for DeepSeek-V4-Pro Preview code something score (lmarena)
68	Gemini 3.5 Flash-LiteGoogle DeepMind	Ranked 68th of 77	1,45395% range 1,446 to 1,460	Behind the leaders	Source for Gemini 3.5 Flash-Lite code something score (lmarena)
69	Gemini 2.5 ProGoogle DeepMind	Ranked 69th of 77	1,45195% range 1,447 to 1,456	Behind the leaders	Source for Gemini 2.5 Pro code something score (lmarena)
70	Inkling-Small Open weightsThinking Machines	Ranked 70th of 77	1,45195% range 1,444 to 1,458	Behind the leaders	Source for Inkling-Small code something score (lmarena)
71	GPT-6 SolOpenAI	Ranked 71st of 77	1,44795% range 1,437 to 1,458	Behind the leaders	Source for GPT-6 Sol code something score (lmarena)
72	Grok 4.20 Beta 1xAI	Ranked 72nd of 77	1,44795% range 1,439 to 1,454	Behind the leaders	Source for Grok 4.20 Beta 1 code something score (lmarena)
73	GPT-6 LunaOpenAI	Ranked 73rd of 77	1,43795% range 1,431 to 1,442	Behind the leaders	Source for GPT-6 Luna code something score (lmarena)
74	GPT-5.5 InstantOpenAI	Ranked 74th of 77	1,43395% range 1,426 to 1,441	Behind the leaders	Source for GPT-5.5 Instant code something score (lmarena)
75	Grok 4.7xAI	Ranked 75th of 77	1,42895% range 1,417 to 1,438	Behind the leaders	Source for Grok 4.7 code something score (lmarena)
76	Grok 4.3xAI	Ranked 76th of 77	1,41495% range 1,409 to 1,420	Behind the leaders	Source for Grok 4.3 code something score (lmarena)
77	Qwen3.5-FlashAlibaba	Ranked 77th of 77	1,41295% range 1,406 to 1,417	Behind the leaders	Source for Qwen3.5-Flash code something score (lmarena)

Test: Arena text: coding. Source: lmarena. Data last changed 29 Oct 2026. How we test.

Want a straight answer for code something?

See the options the evidence supports, with limits and verified prices.

Find my AI