SWE-bench Multilingual (bash only) leaderboard
As of 16 Nov 2026, the top score on SWE-bench Multilingual (bash only) is 72.7% by Gemini 3 Flash (Google), ahead of Claude Opus 4.6 at 72.0%, across 13 model configurations with a published result (SWE-bench, result dated 13 Feb 2026). 300 real GitHub issues from 42 repositories in nine programming languages, run by the SWE-bench team in the minimal bash-only mini-SWE-agent harness. SWE-bench Multilingual (bash only) is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 16 Nov 2026)
- Gemini 3 Flash holds the top SWE-bench Multilingual (bash only) score at 72.7% (SWE-bench, 13 Feb 2026).
- Claude Opus 4.6 is second on SWE-bench Multilingual (bash only) at 72.0% (SWE-bench, 13 Feb 2026).
- SWE-bench Multilingual (bash only): 300 real GitHub issues from 42 repositories in nine programming languages, run by the SWE-bench team in the minimal bash-only mini-SWE-agent harness (SWE-bench).
- 13 model configurations have a published SWE-bench Multilingual (bash only) result on BenchLeader (BenchLeader, 16 Nov 2026).
- SWE-bench Multilingual (bash only) does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
300 real GitHub issues from 42 repositories in nine programming languages, run by the SWE-bench team in the minimal bash-only mini-SWE-agent harness.
- Published by
- SWE-bench
- Category
- Coding
- Index weight
- Reference only
- Models
- 13
- Data as of
- 16 Nov 2026
- Newest result
- 20 Feb 2026
The publisher has posted no new results on this board since 20 Feb 2026. BenchLeader re-reads it every day, so newer models appear here as soon as they are published.
SWE-bench (Jimenez et al. 2024), swebench.com.
- 1Gemini 3 Flash72.7%
- 2Claude Opus 4.672.0%
- 3Claude Opus 4.570.7%
- 4GLM-569.7%
- 5Gemini 3 Pro68.7%
- 6MiniMax M2.568.3%
- 7Kimi K2.567.3%
- 8Claude Sonnet 4.567.0%
- 9GPT-5.2 (high)66.7%
- 10GPT-5.2-Codex66.3%
- 11Claude Haiku 4.564.7%
- 12DeepSeek V3.259.0%
- 13GPT-5 mini39.7%
| # | ||||||
|---|---|---|---|---|---|---|
| 1 | 72.7% | 56.5 | mini-SWE-agent | $105.55 | 2026-02-13 | |
| 2 | 72.0% | 62.1 | mini-SWE-agent | $198.89 | 2026-02-13 | |
| 3 | 70.7% | 57.1 | mini-SWE-agent | $250.07 | 2026-02-13 | |
| 4 | 69.7% | 54.9 | mini-SWE-agent | $192.74 | 2026-02-13 | |
| 5 | 68.7% | 60.0 | mini-SWE-agent | $306.33 | 2026-02-13 | |
| 6 | 68.3% | 53.7 | mini-SWE-agent | $30.06 | 2026-02-16 | |
| 7 | 67.3% | 54.6 | mini-SWE-agent | $207.85 | 2026-02-13 | |
| 8 | 67.0% | 53.8 | mini-SWE-agent | $200.86 | 2026-02-13 | |
| 9 | 66.7% | 55.8 | mini-SWE-agent | $160.92 | 2026-02-13 | |
| 10 | 66.3% | 57.6 | mini-SWE-agent | $198.65 | 2026-02-20 | |
| 11 | 64.7% | 47.7 | mini-SWE-agent | $113.85 | 2026-02-13 | |
| 12 | 59.0% | 50.2 | mini-SWE-agent | $115.15 | 2026-02-13 | |
| 13 | 39.7% | 56.6 | mini-SWE-agent | $15.49 | 2026-02-13 |
AI coding benchmark leaderboard · All LLM benchmarks
Related benchmarks
Cite as: BenchLeader, “SWE-bench Multilingual (bash only) leaderboard”, data as of 16 Nov 2026, https://www.benchleader.com/benchmarks/swe_bench_multilingual.
SWE-bench Multilingual (bash only): questions
- What does SWE-bench Multilingual (bash only) measure?
- 300 real GitHub issues from 42 repositories in nine programming languages, run by the SWE-bench team in the minimal bash-only mini-SWE-agent harness. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on SWE-bench Multilingual (bash only)?
- As of 16 Nov 2026, the highest score on SWE-bench Multilingual (bash only) is 72.7% by Gemini 3 Flash (result dated 13 Feb 2026), ahead of Claude Opus 4.6 at 72.0%. Scores come from SWE-bench and are re-read daily.
- Which AI model leads SWE-bench Multilingual (bash only)?
- Gemini 3 Flash leads SWE-bench Multilingual (bash only) with 72.7% as of 16 Nov 2026 (result dated 13 Feb 2026), ahead of Claude Opus 4.6 at 72.0%.
- How many models have SWE-bench Multilingual (bash only) results?
- 13 model configurations have a SWE-bench Multilingual (bash only) result on BenchLeader, all taken from SWE-bench.
- Who runs SWE-bench Multilingual (bash only) and how often is it updated?
- SWE-bench Multilingual (bash only) is published by SWE-bench. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does SWE-bench Multilingual (bash only) count toward the BenchLeader Index?
- No. SWE-bench Multilingual (bash only) is shown for reference but left out of the composite index.