How is this ranking computed?

It is the same family composite as the leaderboard. Each benchmark is weighted by how much of its spread survives its own measurement error.

Best AI for coding

Ranked by the coding composite: SWE-bench Verified, Aider polyglot and MirrorCode. Full leaderboard.

Rank	Model	Lab	Composite	Evidence	Out $/M
1	o3-pro	—	100th	1 bench · 1 src	$80.00
2	gemini-2.5-pro-preview-06-05	—	99th	1 bench · 1 src	—
3	Claude Opus 5.5	Anthropic	99th	15 bench · 2 src	$20.00
4	Grok 4	xAI	98th	5 bench · 2 src	$15.00
5	gemini-2.5-pro-preview-06-05	—	97th	1 bench · 1 src	—
6	Claude Fable 5.1	Anthropic	96th	14 bench · 2 src	$50.00
7	o3 (high) + gpt-4.1	—	95th	1 bench · 1 src	—
8	Gemini 2.5 Pro Preview 05-06	—	95th	1 bench · 1 src	$10.00
9	GPT 5	OpenAI	94th	13 bench · 3 src	$10.00
10	Gemini 2.5 Pro Preview 03-25	—	93rd	1 bench · 1 src	$10.00
11	Doubao-Seed-Code	ByteDance	92nd	1 bench · 1 src	$1.12
12	GLM-5.2	Zhipu	92nd	10 bench · 2 src	$4.40
13	DeepSeek v4	DeepSeek	91st	7 bench · 1 src	—
14	DeepSeek R1	—	90th	1 bench · 1 src	$2.60
15	Qwen3.7-Max	Alibaba	89th	10 bench · 2 src	$7.50
16	Claude Opus 4.6	Anthropic	89th	4 bench · 1 src	—
17	claude-opus-4-20250514	—	88th	1 bench · 1 src	—
18	Claude 4.5 Opus	Anthropic	87th	1 bench · 1 src	$25.00
19	Qwen3.6 Max Preview	Alibaba	86th	6 bench · 1 src	free
20	Kimi K2.6	Kimi	85th	13 bench · 2 src	$4.00
21	Claude Opus 4.5	Anthropic	85th	6 bench · 1 src	—
22	Claude Fable 5	Anthropic	84th	15 bench · 2 src	$50.00
23	MiniMax M2.5	MiniMax	83rd	1 bench · 1 src	$1.20
24	Gemini 3.1 Pro Preview	Google	82nd	1 bench · 1 src	$12.00
25	Gemini 3 Flash Preview	Google	82nd	10 bench · 2 src	$3.00
26	Claude 4.6 Opus	Anthropic	81st	1 bench · 1 src	—
27	Gemini 3.5 Flash	Google	80th	14 bench · 3 src	$9.00
28	Claude Sonnet 4.6	Anthropic	79th	2 bench · 1 src	—
29	GPT-5.3 Codex	OpenAI	79th	1 bench · 1 src	$14.00
30	Claude Opus 4	Anthropic	78th	7 bench · 2 src	—
31	GLM-5.1	Zhipu	77th	8 bench · 1 src	$4.40
32	Claude Opus 4.1	Anthropic	76th	9 bench · 1 src	$75.00
33	Gemini 3 Pro Preview	Google	75th	5 bench · 2 src	$12.00
34	GPT 5.2 Codex	OpenAI	75th	2 bench · 2 src	$14.00
35	GLM 5	Zhipu	74th	5 bench · 2 src	$3.20
36	GPT 5.2	OpenAI	73rd	13 bench · 3 src	$14.00
37	Kimi K2.5	Kimi	72nd	6 bench · 2 src	$2.80
38	o3	OpenAI	72nd	11 bench · 3 src	$8.00
39	Claude 4.5 Sonnet	Anthropic	71st	1 bench · 1 src	$14.945
40	DeepSeek-V3.2-Exp	DeepSeek	70th	6 bench · 3 src	$0.89
41	Claude Sonnet 4.5	Anthropic	69th	8 bench · 1 src	—
42	DeepSeek R1 + claude-3-5-sonnet-20241022	—	69th	1 bench · 1 src	—
43	Claude 4 Sonnet	Anthropic	68th	1 bench · 1 src	$16.50
44	Claude 4 Opus	Anthropic	67th	1 bench · 1 src	$82.50
45	claude-3-7-sonnet-20250219	—	66th	1 bench · 1 src	—
46	Claude Opus 4.7	Anthropic	65th	14 bench · 2 src	$25.00
47	Qwen3 235B A22B diff, no think, Alibaba API	—	65th	1 bench · 1 src	—
48	Claude Sonnet 4	Anthropic	64th	7 bench · 3 src	—
49	o1-2024-12-17	OpenAI	63rd	9 bench · 3 src	$54.00
50	Claude 4.5 Haiku	Anthropic	62nd	1 bench · 1 src	$5.00
51	GPT 5.1	OpenAI	62nd	8 bench · 2 src	$10.00
52	o4-mini	OpenAI	61st	12 bench · 3 src	$4.00
53	Claude 3.7 Sonnet	Anthropic	60th	7 bench · 3 src	$15.00
54	GPT 5.1 Codex	OpenAI	59th	1 bench · 1 src	$10.00
55	DeepSeek R1	DeepSeek	59th	4 bench · 2 src	$2.60
56	claude-sonnet-4-20250514	—	58th	1 bench · 1 src	—
57	Multiple	Anthropic	57th	1 bench · 1 src	—
58	gemini-2.5-flash-preview-05-20	—	56th	1 bench · 1 src	—
59	DeepSeek V3	—	55th	1 bench · 1 src	—
60	Kimi K2 Instruct	Kimi	55th	6 bench · 3 src	$2.5146
61	Quasar Alpha	—	54th	1 bench · 1 src	—
62	GPT-6 Astra	OpenAI	53rd	15 bench · 2 src	$50.00
63	Claude 3.7 Sonnet w/ Review Heavy	Anthropic	52nd	1 bench · 1 src	—
64	swe-search	Anthropic	52nd	1 bench · 1 src	—
65	Grok 3 Beta	xAI	51st	6 bench · 2 src	—
66	Optimus Alpha	—	50th	1 bench · 1 src	—
67	MiniMax M2	MiniMax	49th	1 bench · 1 src	$1.20
68	GPT 5 mini	OpenAI	49th	11 bench · 2 src	$2.00
69	TTS	OpenAI	48th	1 bench · 1 src	—
70	GPT-5.4	OpenAI	47th	14 bench · 2 src	$15.00
71	GPT-5.5	OpenAI	46th	17 bench · 2 src	$30.00
72	Qwen3.6 Plus	Alibaba	45th	9 bench · 2 src	$3.00
73	DeepSeek Chat V3	—	45th	1 bench · 1 src	—
74	gemini-2.5-flash-preview-04-17	—	44th	1 bench · 1 src	—
75	Devstral Small	Mistral	43rd	1 bench · 1 src	—
76	Qwen3-Coder-30B-A3B-Instruct	Alibaba	42nd	1 bench · 1 src	$0.60
77	Gemini 2.5 Pro	Google	42nd	7 bench · 2 src	$10.00
78	Claude 3.5 Sonnet	Anthropic	41st	2 bench · 2 src	$15.00
79	Qwen3-Coder 480B/A35B Instruct	Alibaba	40th	1 bench · 1 src	$1.80
80	GLM 4.6	Zhipu	39th	3 bench · 2 src	$2.20
81	GPT-4.5	OpenAI	39th	4 bench · 2 src	$150.00
82	GLM 4.5	Zhipu	38th	3 bench · 2 src	$2.20
83	gemini-2.5-flash-preview-05-20	—	37th	1 bench · 1 src	—
84	Devstral	Mistral	36th	1 bench · 1 src	—
85	O3 Mini	OpenAI	35th	12 bench · 3 src	$4.40
86	Frogboss 32B 2510	Microsoft	35th	1 bench · 1 src	—
87	GPT 4.1	OpenAI	34th	10 bench · 3 src	$8.00
88	Grok 3 Mini Beta	—	33rd	1 bench · 1 src	—
89	Qwen3 32B	Alibaba	32nd	3 bench · 2 src	—
90	gemini-exp-1206	—	32nd	1 bench · 1 src	—
91	chatgpt-4o-latest	—	31st	1 bench · 1 src	$14.00
92	TTS	Alibaba	30th	1 bench · 1 src	—
93	DevStral Small 2505	Mistral	29th	1 bench · 1 src	—
94	Gemini 2.0 Pro exp-02-05	—	29th	1 bench · 1 src	$7.956
95	Frogmini 14B 2510	Microsoft	28th	1 bench · 1 src	—
96	o1-mini	OpenAI	27th	5 bench · 2 src	$4.40
97	Amazon.nova Premier v1:0	Amazon	26th	1 bench · 1 src	—
98	DeepSWE-Preview	Agentica	25th	1 bench · 1 src	—
99	GPT-5.6 Sol	OpenAI	25th	13 bench · 2 src	$20.00
100	Llama3-SWE-RL-70B	Meta	24th	1 bench · 1 src	—
101	Claude 3.5 Haiku	Anthropic	23rd	2 bench · 2 src	$4.00
102	SWE-agent-LM-32B	Alibaba	22nd	1 bench · 1 src	—
103	gpt-oss-120b	OpenAI	22nd	5 bench · 3 src	free
104	QwQ-32B + Qwen 2.5 Coder Instruct	—	21st	1 bench · 1 src	—
105	DevStral Small 2507	Mistral	20th	1 bench · 1 src	—
106	GPT 5 nano	OpenAI	19th	11 bench · 2 src	$0.40
107	qwen-max-2025-01-25	—	19th	1 bench · 1 src	$6.40
108	QwQ 32B	Alibaba	18th	3 bench · 2 src	$1.00
109	GPT 4.1 mini	OpenAI	17th	10 bench · 3 src	$1.60
110	Gemini 3.1 Pro Preview	Google	16th	13 bench · 2 src	$12.00
111	Gemini 2.0 Flash Thinking	Google	15th	6 bench · 3 src	$0.42
112	GPT 4o	OpenAI	15th	9 bench · 3 src	$10.00
113	gemini-2.0-flash-thinking-exp-01-21	—	14th	1 bench · 1 src	—
114	Qwen2.5	Alibaba	13th	1 bench · 1 src	—
115	DeepSeek Chat V2.5	—	12th	1 bench · 1 src	—
116	Gemini 2.5 Flash	Google	12th	2 bench · 2 src	$2.50
117	yi-lightning	—	11th	1 bench · 1 src	—
118	command-a-03-2025-quality	—	10th	1 bench · 1 src	—
119	Codestral 25.01	—	9th	1 bench · 1 src	—
120	Llama 4 Maverick Instruct	Meta	9th	6 bench · 3 src	$0.60
121	Qwen2.5 Coder 32B Instruct	Alibaba	8th	2 bench · 2 src	$1.00
122	openhands-lm-32b-v0.1	—	7th	1 bench · 1 src	—
123	GPT-4.1 nano	OpenAI	6th	7 bench · 2 src	$0.40
124	MCTS Refine 7B	—	5th	1 bench · 1 src	—
125	Gemma 3 27B	Google	5th	4 bench · 2 src	free
126	GPT-4o mini	OpenAI	4th	8 bench · 2 src	$0.60
127	GPT-4	OpenAI	3rd	1 bench · 1 src	—
128	Claude 3 Opus	Anthropic	2nd	6 bench · 2 src	$75.00
129	Llama 4 Scout Instruct	Meta	2nd	5 bench · 2 src	$0.30
130	Claude 2	Anthropic	1st	4 bench · 2 src	—
131	SWE-Llama 7B	Meta	0th	1 bench · 1 src	—

Common questions

What is the best ai for coding right now?: Ranked by the coding composite: SWE-bench Verified, Aider polyglot and MirrorCode.
How is this ranking computed?: It is the same family composite as the leaderboard. Each benchmark is weighted by how much of its spread survives its own measurement error.