SWE-bench Verified (Epoch) leaderboard
As of 7 Nov 2026, the top score on SWE-bench Verified (Epoch) is 83.5% by Claude Opus 4.7 (Anthropic), ahead of GPT-5.5 Pre Release at 80.6%, across 33 model configurations with a published result (Epoch AI Benchmarking Hub, result dated 16 Apr 2026). SWE-bench Verified is 500 real GitHub issues from popular Python repositories, each screened by human annotators to make sure the task is well specified and the tests are fair. The model reads the issue, edits the repository, and its patch is accepted only if the hidden tests pass. SWE-bench Verified (Epoch) counts toward the BenchLeader coding score.
Key facts (as of 7 Nov 2026)
- Claude Opus 4.7 holds the top SWE-bench Verified (Epoch) score at 83.5% (Epoch AI Benchmarking Hub, 16 Apr 2026).
- GPT-5.5 Pre Release is second on SWE-bench Verified (Epoch) at 80.6% (Epoch AI Benchmarking Hub, 23 Apr 2026).
- SWE-bench Verified (Epoch): Epoch AI's own SWE-bench Verified runs with a constant scaffold (Epoch AI Benchmarking Hub).
- 33 model configurations have a published SWE-bench Verified (Epoch) result on BenchLeader (BenchLeader, 7 Nov 2026).
- SWE-bench Verified (Epoch) feeds the coding category of the BenchLeader Index (BenchLeader methodology).
Epoch AI's own SWE-bench Verified runs with a constant scaffold.
- Published by
- Epoch AI Benchmarking Hub
- Category
- Coding
- Index weight
- 0.5
- Models
- 33
- Data as of
- 7 Nov 2026
CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.
What the test looks like
SWE-bench Verified is 500 real GitHub issues from popular Python repositories, each screened by human annotators to make sure the task is well specified and the tests are fair. The model reads the issue, edits the repository, and its patch is accepted only if the hidden tests pass.
How it is scored
Percent of the 500 issues resolved, run by Epoch AI.
What to keep in mind
Epoch AI's own SWE-bench Verified runs, with one constant scaffold across every model. A fixed scaffold makes models comparable but can understate those tuned for a different agent loop. Numbers are usually below the labs' own best reports. See also the SWE-bench Verified (bash only) board.
- 1Claude Opus 4.7 (max)83.5%
- 2GPT-5.5 Pre Release (xhigh)80.6%
- 3Gemini 3.5 Flash (high)79.3%
- 4Claude Opus 4.678.7%
- 5GLM 5.2 (max)78.7%
- 6DeepSeek V4 Pro (max)77.6%
- 7Qwen3.7 (max)77.3%
- 8GPT-5.4 (high)76.9%
- 9Qwen3.6 Max (max)76.7%
- 10Claude Opus 4.576.7%
- 11Kimi K2.676.7%
- 12Gemini 3.1 Pro Customtools75.6%
- 13Gemini 3 Flash75.4%
- 14Claude Sonnet 4.675.2%
- 15GPT-5.3-Codex (high)74.8%
| # | ||||
|---|---|---|---|---|
| 1 | 83.5% | 59.5 | 2026-04-16 | |
| 2 | 80.6% | – | 2026-04-23 | |
| 3 | 79.3% | 62.5 | 2026-05-19 | |
| 4 | 78.7% | 62.3 | 2026-02-05 | |
| 5 | 78.7% | 59.5 | 2026-06-16 | |
| 6 | 77.6% | 57.0 | 2026-04-24 | |
| 7 | 77.3% | 60.3 | 2026-05-19 | |
| 8 | 76.9% | 58.8 | 2026-03-05 | |
| 9 | 76.7% | 62.4 | 2026-04-20 | |
| 10 | 76.7% | 57.9 | 2025-11-24 | |
| 11 | 76.7% | 55.2 | 2026-04-20 | |
| 12 | 75.6% | – | 2026-02-19 | |
| 13 | 75.4% | 56.9 | 2025-12-17 | |
| 14 | 75.2% | 58.2 | 2026-02-17 | |
| 15 | 74.8% | – | 2026-02-05 | |
| 16 | 74.2% | 55.7 | 2026-04-07 | |
| 17 | 73.8% | 55.9 | 2025-12-11 | |
| 18 | 73.8% | 54.2 | 2026-01-27 | |
| 19 | 73.5% | 57.9 | 2025-08-07 | |
| 20 | 73.3% | 51.4 | 2025-08-05 | |
| 21 | 72.9% | 60.2 | 2025-11-18 | |
| 22 | 72.1% | 54.8 | 2026-02-11 | |
| 23 | 71.5% | 58.1 | 2025-08-07 | |
| 24 | 71.3% | 53.7 | 2025-09-29 | |
| 25 | 70.7% | 52.6 | 2025-05-22 | |
| 26 | 68.0% | 57.7 | 2025-11-13 | |
| 27 | 64.7% | 53.5 | 2025-08-07 | |
| 28 | 62.3% | 53.3 | 2025-04-16 | |
| 29 | 61.0% | 49.9 | 2025-02-24 | |
| 30 | 57.9% | 59.1 | 2026-03-31 | |
| 31 | 57.6% | 54.5 | 2025-06-17 | |
| 32 | 48.5% | 48.5 | 2025-04-14 | |
| 33 | 31.0% | 43.3 | 2024-11-20 |
More coding benchmarks · All LLM benchmarks
Related benchmarks
Top head-to-heads on SWE-bench Verified (Epoch)
Cite as: BenchLeader, “SWE-bench Verified (Epoch) leaderboard”, data as of 7 Nov 2026, https://www.benchleader.com/benchmarks/swe_bench_verified_epoch.
SWE-bench Verified (Epoch): questions
- What does SWE-bench Verified (Epoch) measure?
- SWE-bench Verified is 500 real GitHub issues from popular Python repositories, each screened by human annotators to make sure the task is well specified and the tests are fair. The model reads the issue, edits the repository, and its patch is accepted only if the hidden tests pass. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on SWE-bench Verified (Epoch)?
- As of 7 Nov 2026, the highest score on SWE-bench Verified (Epoch) is 83.5% by Claude Opus 4.7 (result dated 16 Apr 2026), ahead of GPT-5.5 Pre Release at 80.6%. Scores come from Epoch AI Benchmarking Hub and are re-read daily.
- Which AI model leads SWE-bench Verified (Epoch)?
- Claude Opus 4.7 leads SWE-bench Verified (Epoch) with 83.5% as of 7 Nov 2026 (result dated 16 Apr 2026), ahead of GPT-5.5 Pre Release at 80.6%.
- How many models have SWE-bench Verified (Epoch) results?
- 33 model configurations have a SWE-bench Verified (Epoch) result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
- Who runs SWE-bench Verified (Epoch) and how often is it updated?
- SWE-bench Verified (Epoch) is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does SWE-bench Verified (Epoch) count toward the BenchLeader Index?
- Yes. SWE-bench Verified (Epoch) contributes to the coding category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.