Best AI models for coding

Ranked by SWE-bench Verified: the share of real bugs from real open-source projects that each model fixes on its own. Tested independently by Epoch AI, shown with current API prices.

Updated September 30, 2026

Our picks

#ModelSWE-bench VerifiediOverall indexiInputiOutputiContexti
1
83.5%
156.4$5$251M
2
Gemini 3.5 FlashGoogle DeepMind
79.3%
154.6$1.50$91M
3
GLM 5.2Z.ai (Zhipu)
78.7%
151.8$0.14$3.991M
4
78.7%
155.4$5$251M
5
77.6%
155.4$0.66$1.981M
6
Qwen3.7 MaxAlibaba Qwen
77.3%
153.7$1.48$4.431M
7
GPT-5.4OpenAI
76.9%
156.9$2.50$151.1M
8
76.7%
149.3$1.03$6.16256K
9
Kimi K2.6Moonshot AI
76.7%
151.0$0.65$3.41256K
10
76.7%
150.1$5$25200K

per 1M tokens · USD

Quick answers

Which AI model is best for coding?
Claude Opus 4.7 from Anthropic leads this ranking (SWE-bench Verified: 83.5%). It costs $5 per million input tokens and $25 per million output tokens.
Is there a cheaper alternative?
DeepSeek V4 Pro 0813 from DeepSeek (SWE-bench Verified: 77.6%) costs $0.66 / $1.98 per million tokens, about 10× less than Claude Opus 4.7.

How we rank

Scores come from tests run independently by Epoch AI (CC BY 4.0), not from the companies’ own announcements. Prices come from OpenRouter’s public API. Both are refreshed automatically several times a day, so this ranking follows new releases and price changes.

Other rankings

MathScienceCheapFreeOpen sourceLong contextImages & PDFsRankings →