SWE-bench Multilingual leaderboard
- Claude Opus 4.6 leads
- 72.0%
- models measured
- 11
- lead over #2
- +1.3 pts
- median
- 67.0%
Leader on SWE-bench Multilingual · read October 22
# SWE-bench Multilingual, read October 22
leaders = {
"Claude Opus 4.6": 72.0%, "Claude Opus 4.5": 70.7%, "GLM 5": 69.7%,
}
On October 22, 2026, 11 of our catalog's models had a SWE-bench Multilingual result, and Anthropic Claude Opus 4.6 led with 72.0%, ahead of Anthropic Claude Opus 4.5 at 70.7%.
GitHub issues from projects in languages other than Python, graded the SWE-bench way: the patch has to make the project tests pass. The number is the share of issues resolved, as swebench.com publishes it.
Role in the index
A coding board we show beside the Vibe Coding Index, not inside it: the index keeps the six boards it was defined on, so a score does not move every time a board is added. Read the methodology
| # | Model | Result | Cost per attempt | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.6Anthropic | 72.0% | $0.663 | 57.2 | $25.00 | |
| 2 | Claude Opus 4.5Anthropic | 70.7% | $0.8336 | 44.0 | $25.00 | |
| 3 | GLM 5Z.ai | 69.7% | $0.6425 | 41.7 | $3.20 | |
| 4 | MiniMax M2.5MiniMax | 68.3% | $0.1012 | 37.2 | $1.20 | |
| 5 | Kimi K2.5Moonshot AI | 67.3% | $0.6928 | 38.6 | $2.85 | |
| 6 | Claude Sonnet 4.5Anthropic | 67.0% | $0.6695 | 39.5 | $15.00 | |
| 7 | GPT-5.2OpenAI | 66.7% | $0.5364 | 55.9 | $14.00 | |
| 8 | GPT-5.2-CodexOpenAI | 66.3% | $0.6622 | 45.4 | $14.00 | |
| 9 | Claude Haiku 4.5Anthropic | 64.7% | $0.3795 | 35.6 | $5.00 | |
| 10 | DeepSeek V3.2DeepSeek | 59.0% | $0.3838 | 30.0 | $0.88 | |
| 11 | GPT-5 MiniOpenAI | 39.7% | $0.0516 | 34.2 | $2.00 |
Only models in our catalog are listed, and each appears once with its best published run. 42 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Cost per attempt is what swebench.com reports one task cost on the run behind the result, in US dollars. Only some runs report it.