// BENCHMARK

SWE-Bench Multilingual benchmark

AI model leaderboard for the SWE-Bench Multilingual benchmark. Compare how large language models score on SWE-Bench Multilingual, see the full ranking, and understand what this AI benchmark measures. DeepSeek-V3.2 currently leads with 70.2. Multilingual extension of SWE-Bench — real GitHub issues across Python, Java, JavaScript, Go, etc.

Leaderboard

# Model Organization Score Variant Source
#1 DeepSeek-V3.2 DeepSeek 70.2 cited-devstral-readme official ↗
#2 DeepSeek-V3.2 DeepSeek 70.2 thinking official ↗
#3 Claude Sonnet 4.5 Anthropic 68 cited-devstral-readme official ↗
#4 Claude Sonnet 4.5 Anthropic 68 cited-deepseek-v3-2 official ↗
#5 Devstral 2 Mistral AI 61.3 from-hf-readme official ↗
#6 Kimi K2 Thinking Moonshot 61.1 thinking-w-tools official ↗
#7 Kimi K2 Thinking Moonshot 61.1 cited-devstral-readme official ↗
#8 Kimi K2 Thinking Moonshot 61.1 thinking-cited-deepseek-v3-2 official ↗
#9 DeepSeek-V3.2-Exp DeepSeek 57.9 from-hf-readme official ↗
#10 MiniMax-M2 MiniMax 56.5 cited-deepseek-v3-2 official ↗
#11 MiniMax-M2 MiniMax 56.5 cited-devstral-readme official ↗
#12 Kimi K2 Moonshot 55.9 k2-0905-w-tools-cited-k2-thinking official ↗
#13 Devstral Small 2 Mistral AI 55.7 from-hf-readme official ↗
#14 GPT-5 OpenAI 55.3 cited-deepseek-v3-2 official ↗

Frequently asked questions about SWE-Bench Multilingual

What is the SWE-Bench Multilingual benchmark?

Multilingual extension of SWE-Bench — real GitHub issues across Python, Java, JavaScript, Go, etc.

How is the SWE-Bench Multilingual benchmark scored?

SWE-Bench Multilingual is scored using the resolution_rate metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on SWE-Bench Multilingual?

As of the latest reported scores on GenAIList, DeepSeek-V3.2 achieves the highest result on SWE-Bench Multilingual with a score of 70.2.

Is a higher SWE-Bench Multilingual score better?

Yes. On SWE-Bench Multilingual a higher score indicates better performance, so models near the top of the leaderboard are the strongest.