SWE-bench Pro leaderboard

A long-horizon repository benchmark built to test realistic software engineering work. Its scores need a task-quality and setup check before they support a coding-agent decision.

Models ranked by their published SWE-bench Pro value

Rank	Model	Provider	Repository task completion
1	Claude Fable 5.1	Anthropic	81.2
2	Claude Mythos 5	Anthropic	80.3
3	Claude Fable 5	Anthropic	80
4	Claude Opus 5	Anthropic	79.2
5	SASakana Fugu-Ultra	Sakana AI	73.7
6	Claude Opus 4.8	Anthropic	69.2
7	Qwen3.8 Max	Alibaba	67.7
8	Hy4 preview	Tencent	65.7
9	OAOrnith-1.5-397B	Ornith AI	65.1
10	Grok 4.5	xAI	64.7
11	GPT-5.6 Sol	OpenAI	64.6
12	Claude Opus 4.7 (Adaptive)	Anthropic	64.3
13	GPT-5.6 Terra	OpenAI	63.4
14	Qwen3.8-Omni-Flash	Alibaba	63.3
15	Claude Sonnet 5	Anthropic	63.2
16	GPT-5.6 Luna	OpenAI	62.7
17	Qwen3.8-Flash-Next	Alibaba	62.5
18	DAOrnith-1.0-397B	DeepReinforce AI	62.2
19	GLM-5.2	Z.AI	62.1
20	Qwen3.8-27B	Alibaba	61.7
21	Muse Spark 1.1	Meta	61.5
22	DSdots3-note Preview	Dots Studio	61
23	Qwen3.7 Max	Alibaba	60.6
24	Atria Dawn Preview	Shanghai Artificial Intelligence Laboratory	59.6
24	OAOrnith-1.5-35B-A3B	Ornith AI	59.6
26	Laguna S 2.1	Poolside	59.4
27	MiniMax M3	MiniMax	59
27	SASakana Fugu	Sakana AI	59
29	GPT-5.5	OpenAI	58.6
29	Kimi K2.6	Moonshot AI	58.6
31	GLM-5.1	Z.AI	58.4
32	GPT-5.4	OpenAI	57.7
33	Qwen3.7 Plus	Alibaba	57.6
34	Qwen 3.6 Max (preview)	Alibaba	57.3
35	MiMo-V2.5-Pro	Xiaomi	57.2
36	Claude Opus 4.5	Anthropic	57.1
37	GPT-5.3 Codex	OpenAI	56.8
38	Ling 3.0 Flash	InclusionAI	56.6
38	Qwen3.6 Plus	Alibaba	56.6
40	Step 3.7 Flash	StepFun	56.3
41	MiniMax M2.7	MiniMax	56.2
42	MiMo-V2.5	Xiaomi	56.1
43	TMInkling-Small	Thinking Machines Lab	55.9
44	GPT-5.2	OpenAI	55.6
45	DeepSeek V4 Pro 0813	DeepSeek	55.4
46	Gemini 3.5 Flash	Google	55.1
46	GLM-5	Z.AI	55.1
48	TMInkling	Thinking Machines Lab	54.3
49	Gemini 3.5 Flash-Lite	Google	54.2
50	Qwen3.6-27B	Alibaba	53.5
51	Claude Opus 4.6	Anthropic	53.4
52	MAI-Thinking-1	Microsoft	52.8
53	DeepSeek V4 Flash 0731	DeepSeek	52.6
54	Muse Spark	Meta	52.4
55	Grok 4.20	xAI	51.8
56	Muse Glimmer 30B	Meta	51.2
57	Qwen3.5 397B	Alibaba	50.9
58	Kimi K2.5	Moonshot AI	50.7
59	DAOrnith-1.0-35B	DeepReinforce AI	50.4
60	Qwen3.6-35B-A3B	Alibaba	49.5
61	Laguna M.1	Poolside	49.2
62	Laguna XS 2.1	Poolside	47.6
63	OAOrnith-1.5-9B	Ornith AI	47.5
64	Laguna XS.2	Poolside	46.3
65	DAOrnith-1.0-9B	DeepReinforce AI	42.9
66	LongCat-Flash-Lite-Sparse	Meituan	40.63
67	Granite 4.2 30B	IBM	33.29
68	LLaDA2.2-flash	InclusionAI	30.1
69	Granite 4.2 8B	IBM	19.11
70	OPMiniCPM5-2B	OpenBMB	14.4

Evidence key: Observed

Rows are ordered by the value SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks? published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.