Best AI Models for Coding

Q: Which AI model is best for coding?

The top-ranked coding model changes frequently. Check the live leaderboard above for the current leader based on HumanEval+, SWE-bench Verified, and other coding benchmarks.

Q: What benchmarks measure coding ability?

Key coding benchmarks include HumanEval+ (function completion), SWE-bench Verified (real GitHub issues), Aider Polyglot (multi-language editing), and Terminal-bench (CLI tasks).

Q: Are open-source models competitive for coding?

Yes. Several open-source models score competitively on coding benchmarks, especially models from DeepSeek and Meta.

AI models ranked by coding benchmarks. Compare HumanEval+, SWE-bench Verified, Aider Polyglot, and more across all providers.

104

Models

Providers

Open Source

$0.95

Median $/1M in

Top 3

Claude Mythos Preview

🇺🇸 Anthropic

1.0MN/A/M in

81.814 benchmarks

Full Rankings

#	Model	Avg	aider poly	GSO-Bench	swe bench	swe bench	terminal b	$/1M in	Context
1	GPT-5.5🇺🇸 OpenAI	85.0	-	-	-	-	82.7	$5.00	400K
2	GPT-5 Chat🇺🇸 OpenAI	81.9	88.0	-	-	-	-	$1.25	128K
3	Claude Mythos Preview🇺🇸 Anthropic	81.8	-	-	93.9	-	82.0	N/A	1.0M
4	Gemini 2.5 Pro Preview 05-06🇺🇸 Google DeepMind	76.9	76.9	-	-	-	-	$1.25	1.0M
5	o4 Mini High🇺🇸 OpenAI	72.0	72.0	-	-	-	-	$1.10	200K
6	Grok 3 Beta🇺🇸 xAI	69.5	53.3	-	-	-	-	$3.00	131K
7	gpt-oss-120b (free)🇺🇸 OpenAIOpen	68.7	41.8	-	-	-	-	Free	131K
8	Grok 3 Mini Beta🇺🇸 xAI	64.8	49.3	-	-	-	-	$0.30	131K
9	o3 Pro🇺🇸 OpenAI	61.3	84.9	-	-	-	-	$20.00	200K
10	Qwen3.7 Max🇨🇳 Alibaba QwenOpen	61.3	-	-	77.3	-	-	$1.25	1.0M
11	o3 Mini High🇺🇸 OpenAI	60.4	60.4	-	-	-	-	$1.10	200K
12	R1 0528🇨🇳 DeepSeekOpen	57.9	71.4	-	-	-	-	$0.50	164K
13	Claude Sonnet 5🇺🇸 Anthropic	56.7	-	37.3	-	-	-	$2.00	1.0M
14	GLM 5.2🇨🇳 z-aiOpen	56.4	-	-	78.7	-	-	$1.00	1.0M
15	Gemini 3 Pro🇺🇸 Google DeepMind	56.1	-	18.6	72.9	-	69.4	N/A	0K
16	DeepSeek V3🇨🇳 DeepSeekOpen	56.1	48.4	-	-	-	-	$0.26	164K
17	Kimi K2 0711🇨🇳 moonshotaiOpen	56.1	59.1	4.9	-	-	27.8	$0.57	131K
18	Claude Opus 4.8🇺🇸 Anthropic	55.4	-	47.1	-	-	-	$5.00	1.0M
19	GPT-5.3-Codex🇺🇸 OpenAI	55.4	-	-	74.8	-	77.3	$1.75	400K
20	DeepSeek V3 0324🇨🇳 DeepSeekOpen	55.1	55.1	-	-	-	-	$0.25	164K
21	Gemini 3.5 Flash🇺🇸 Google DeepMind	54.8	-	-	79.3	-	-	$1.50	1.0M
22	GLM 5🇨🇳 z-aiOpen	54.7	-	-	72.1	-	52.4	$0.60	205K
23	GLM 5.1🇨🇳 z-aiOpen	54.3	-	-	74.2	-	-	$0.97	205K
24	DeepSeek V4 Pro🇨🇳 DeepSeekOpen	53.9	-	-	77.6	-	-	$0.21	1.0M
25	Grok 4🇺🇸 xAI	53.9	79.6	-	-	-	27.2	N/A	0K
26	Qwen3 235B A22B🇨🇳 Alibaba QwenOpen	53.7	59.6	-	-	-	-	$0.46	131K
27	Gemini 2.0 Pro🇺🇸 Google DeepMind	53.7	35.6	-	-	-	-	N/A	0K
28	Grok-3 mini🇺🇸 xAI	53.6	49.3	-	-	-	-	N/A	0K
29	Qwen3.6 Plus🇨🇳 Alibaba QwenOpen	53.5	-	-	57.9	-	-	$0.33	1.0M
30	Grok 4 Fast🇺🇸 xAI	53.3	-	-	-	-	-	N/A	0K
31	Qwen2.5 Coder 32B Instruct🇨🇳 Alibaba QwenOpen	53.1	16.4	-	-	-	-	$0.66	33K
32	gpt-oss-20b🇺🇸 OpenAIOpen	53.1	-	-	-	-	3.4	$0.02	131K
33	GPT-5🇺🇸 OpenAI	53.1	88.0	6.9	73.5	65.0	49.6	$1.25	400K
34	Kimi K2 Thinking🇨🇳 moonshotaiOpen	52.2	-	-	-	63.4	35.7	$0.60	262K
35	GPT-5.4🇺🇸 OpenAI	51.7	-	31.4	76.9	-	81.8	$2.50	1.1M
36	Qwen3 32B🇨🇳 Alibaba QwenOpen	51.5	40.0	-	-	-	-	$0.08	131K
37	GPT-5 Mini🇺🇸 OpenAI	51.5	-	-	64.7	59.8	34.8	$0.25	400K
38	o3🇺🇸 OpenAI	51.2	81.3	8.8	62.3	58.4	-	$2.00	200K
39	Qwen3.6 Max Preview🇨🇳 Alibaba QwenOpen	51.0	-	-	76.7	-	-	$1.03	262K
40	Gemini 2.5 Pro Preview 06-05🇺🇸 Google DeepMind	50.9	83.1	-	-	-	-	$1.25	1.0M
41	Gemini 2.5 Pro🇺🇸 Google DeepMind	50.8	83.1	3.9	57.6	-	32.6	$1.25	1.0M
42	GLM 4.6🇨🇳 z-aiOpen	50.8	-	-	-	-	24.5	$0.43	205K
43	Claude Opus 4.7🇺🇸 Anthropic	50.5	-	44.1	83.5	-	80.2	$5.00	1.0M
44	MiniMax M2.7🇨🇳 minimaxOpen	50.5	-	-	-	-	45.1	$0.21	205K
45	Gemini 3 Flash Preview🇺🇸 Google DeepMind	49.6	-	9.8	75.4	-	64.3	$0.50	1.0M
46	Kimi K2.5🇨🇳 moonshotaiOpen	49.4	-	-	73.8	-	43.2	$0.45	262K
47	gpt-oss-120b🇺🇸 OpenAIOpen	49.3	41.8	-	-	26.0	18.7	$0.04	131K
48	Grok 4.20🇺🇸 xAI	48.8	-	-	-	-	57.3	$1.25	2.0M
49	MiniMax M2.5🇨🇳 minimaxOpen	48.6	-	-	-	-	42.7	$0.27	205K
50	o4 Mini🇺🇸 OpenAI	48.6	72.0	3.6	-	45.0	-	$1.10	200K
51	DeepSeek V3.2 Exp🇨🇳 DeepSeekOpen	48.4	74.2	-	-	-	-	$0.27	164K
52	Qwen3 235B A22B Instruct 2507🇨🇳 Alibaba QwenOpen	48.3	59.6	-	-	-	-	$0.09	262K
53	Gemini 3.1 Pro Preview🇺🇸 Google DeepMind	48.2	-	22.6	75.6	-	80.2	$2.00	1.0M
54	Claude 3.7 Sonnet🇺🇸 Anthropic	48.1	64.9	3.8	61.0	52.8	-	N/A	0K
55	Claude Opus 4.6🇺🇸 Anthropic	48.1	-	33.3	78.7	-	79.8	$5.00	1.0M
56	Gemini 2.0 Flash🇺🇸 Google DeepMind	48.0	38.2	-	-	-	-	$0.10	1.0M
57	DeepSeek V3.2🇨🇳 DeepSeekOpen	47.7	74.2	-	-	-	39.5	$0.28	164K
58	Gemini 2.0 Flash (Dec 2024)🇺🇸 Google DeepMind	47.6	38.2	-	-	-	-	N/A	0K
59	GPT-5.2🇺🇸 OpenAI	47.0	-	27.4	73.8	71.8	64.9	$1.75	400K
60	GPT-5.1🇺🇸 OpenAI	47.0	-	13.7	68.0	66.0	47.6	$1.25	400K
61	Claude Opus 4🇺🇸 Anthropic	46.1	72.0	6.9	70.7	67.6	-	$15.00	200K
62	GLM 4.7🇨🇳 z-aiOpen	46.0	-	-	-	-	33.4	$0.60	205K
63	R1🇨🇳 DeepSeekOpen	45.6	56.9	-	-	-	-	$0.70	64K
64	Claude Sonnet 4🇺🇸 Anthropic	45.5	61.3	4.9	-	64.9	-	$3.00	200K
65	Grok 3🇺🇸 xAI	45.5	53.3	-	-	-	-	N/A	0K
66	o1🇺🇸 OpenAI	45.3	61.7	-	-	-	-	$15.00	200K
67	Claude Sonnet 4.6🇺🇸 Anthropic	45.3	-	-	75.2	-	53.4	$3.00	1.0M
68	GPT-4o-mini (2024-07-18)🇺🇸 OpenAI	43.2	3.6	-	-	-	-	$0.15	128K
69	Gemma 3 27B (free)🇺🇸 Google DeepMindOpen	42.2	4.9	-	-	-	-	Free	131K
70	o1-preview🇺🇸 OpenAI	41.9	-	-	-	-	-	N/A	0K
71	Claude Opus 4.5🇺🇸 Anthropic	41.6	-	26.5	76.7	74.4	63.1	$5.00	200K
72	Qwen3.6 35B A3B🇨🇳 Alibaba QwenOpen	41.6	-	-	-	-	23.0	$0.15	262K
73	Gemini 1.5 Pro (Feb 2024)🇺🇸 Google DeepMind	41.3	-	-	-	-	-	N/A	0K
74	Qwen2.5-Max🇨🇳 Alibaba QwenOpen	41.1	21.8	-	-	-	-	N/A	0K
75	GPT-4.1🇺🇸 OpenAI	39.5	52.4	-	48.5	39.6	-	$2.00	1.0M
76	Claude 3.5 Sonnet🇺🇸 Anthropic	39.4	51.6	4.6	-	-	-	N/A	0K
77	Gemini 2.5 Flash🇺🇸 Google DeepMind	39.3	47.1	-	-	-	17.1	$0.30	1.0M
78	GPT-5 Nano🇺🇸 OpenAI	39.2	-	-	-	34.8	21.8	$0.05	400K
79	Qwen3.5-9B🇨🇳 Alibaba QwenOpen	38.5	-	-	-	-	9.2	$0.10	262K
80	Gemini 1.5 Pro (May 2024)🇺🇸 Google DeepMind	38.4	-	-	-	-	-	N/A	0K
81	Kimi K2.6🇨🇳 moonshotaiOpen	38.3	-	-	76.7	-	-	$0.95	262K
82	Llama 3.1 405B🇺🇸 MetaOpen	37.9	-	-	-	-	-	N/A	0K
83	Claude Sonnet 4.5🇺🇸 Anthropic	37.8	-	14.7	71.3	70.6	46.5	$3.00	1.0M
84	Gemini 2.0 Flash Thinking (Jan 2025)🇺🇸 Google DeepMind	37.7	18.2	-	-	-	-	N/A	0K
85	Claude 3.5 Haiku🇺🇸 Anthropic	36.7	28.0	-	-	-	-	N/A	0K
86	GPT-4.1 Mini🇺🇸 OpenAI	36.6	32.4	-	-	23.9	-	$0.40	1.0M
87	Gemma 3 27B🇺🇸 Google DeepMindOpen	36.6	4.9	-	-	-	-	$0.08	131K
88	Claude Opus 4.1🇺🇸 Anthropic	36.3	-	-	73.3	-	38.0	$15.00	200K
89	GPT-4.5🇺🇸 OpenAI	35.9	44.9	-	-	-	-	N/A	0K
90	GPT-4o (2024-08-06)🇺🇸 OpenAI	35.6	23.1	-	-	-	-	$2.50	128K
91	o1-mini🇺🇸 OpenAI	35.0	32.9	-	-	-	-	N/A	0K
92	Claude Haiku 4.5🇺🇸 Anthropic	33.6	-	-	-	-	35.5	$1.00	200K
93	o3 Mini🇺🇸 OpenAI	32.5	60.4	1.3	-	-	-	$1.10	200K
94	Llama 3 70B Instruct🇺🇸 MetaOpen	32.4	-	-	-	-	-	$0.51	8K
95	GPT-4o (2024-11-20)🇺🇸 OpenAI	32.3	23.1	0.0	31.0	21.6	-	$2.50	128K
96	Llama 3-70B🇺🇸 MetaOpen	32.1	-	-	-	-	-	N/A	0K
97	QwQ 32B🇨🇳 Alibaba QwenOpen	30.9	20.9	-	-	-	-	$0.15	131K
98	GPT-4.1 Nano🇺🇸 OpenAI	30.9	8.9	-	-	-	-	$0.10	1.0M
99	GPT-4o-mini🇺🇸 OpenAI	29.2	3.6	-	-	-	-	$0.15	128K
100	Claude 3 Opus🇺🇸 Anthropic	28.0	-	-	-	-	-	N/A	0K
101	Llama 4 Maverick🇺🇸 MetaOpen	27.9	15.6	-	-	21.0	-	$0.19	1.0M
102	Claude 3 Haiku🇺🇸 Anthropic	25.7	-	-	-	-	-	$0.25	200K
103	Mixtral 8x22B Instruct🇫🇷 Mistral AIOpen	23.8	-	-	-	-	-	$2.00	66K
104	Llama 4 Scout🇺🇸 MetaOpen	19.5	-	-	-	9.1	-	$0.10	1.3M

90+ Gold 80-89 70-79 60-69 <60Scores in % unless noted. Avg = unweighted mean across tested benchmarks.

About this category

Models ranked by coding performance across HumanEval+, SWE-bench Verified, Aider Polyglot, and other code generation benchmarks. Scores reflect real-world software engineering tasks including bug fixing, code completion, and multi-file editing.

Related categories

Best AI Models for Reasoning

AI models ranked by reasoning benchmarks. Compare GPQA Diamond, ARC-AGI, BBH, and other reasoning tests across all providers.

Best AI Models for Math

AI models ranked by math benchmarks. Compare MATH-500, GSM8K, and competition-level math scores across all providers.

Flagship AI Models

The best AI model from each provider, ranked by benchmark score. Compare the flagships from OpenAI, Anthropic, Google, Meta, and more.

Frequently asked questions

Which AI model is best for coding?

The top-ranked coding model changes frequently. Check the live leaderboard above for the current leader based on HumanEval+, SWE-bench Verified, and other coding benchmarks.

What benchmarks measure coding ability?

Key coding benchmarks include HumanEval+ (function completion), SWE-bench Verified (real GitHub issues), Aider Polyglot (multi-language editing), and Terminal-bench (CLI tasks).

Are open-source models competitive for coding?

Yes. Several open-source models score competitively on coding benchmarks, especially models from DeepSeek and Meta.

Back to all models