SWE-bench (Vals) leaderboard
As of 15 Nov 2026, the top score on SWE-bench (Vals) is 97.0% by Claude Opus 5 (Anthropic), ahead of DeepSeek V4 Pro at 96.4%, across 86 model configurations with a published result (Vals AI, result dated 1 Sep 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 (Vals) is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 15 Nov 2026)
- Claude Opus 5 holds the top SWE-bench (Vals) score at 97.0% (Vals AI, 1 Sep 2026).
- DeepSeek V4 Pro is second on SWE-bench (Vals) at 96.4% (Vals AI, 1 Sep 2026).
- SWE-bench (Vals): Vals AI's own SWE-bench run. Shown for reference (Vals AI).
- 86 model configurations have a published SWE-bench (Vals) result on BenchLeader (BenchLeader, 15 Nov 2026).
- SWE-bench (Vals) does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Vals AI's own SWE-bench run. Shown for reference.
- Published by
- Vals AI
- Category
- Coding
- Index weight
- Reference only
- Models
- 86
- Data as of
- 15 Nov 2026
Vals AI (vals.ai).
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 in Vals AI's own agent harness, run by Vals AI, not the bash-only mini-SWE-agent harness of the official board.
What to keep in mind
Vals AI's own run of SWE-bench Verified. Shown for reference; the official bash-only board is counted in the index. See also the SWE-bench Verified (bash only) board.
- 1Claude Opus 597.0%
- 2DeepSeek V4 Pro (max)96.4%
- 3GPT-5.6 Sol (max)96.2%
- 4Grok 4.6 (high)95.6%
- 5GPT-5.6 Terra (max)95.4%
- 6GLM 5.3 (max)95.4%
- 7Claude Fable 5 (max)95.0%
- 8Kimi K393.4%
- 9GPT-5.6 Luna (max)93.0%
- 10GLM 5.3 Flash (max)92.0%
- 11DeepSeek V4 Flash (high)88.8%
- 12Claude Opus 4.8 (max)88.6%
- 13Muse Spark 1.2 (xhigh)86.6%
- 14Grok 4.5 (high)86.6%
- 15Qwen3.8 27B (xhigh)86.0%
| # | |||||
|---|---|---|---|---|---|
| 1 | 97.0% | 63.2 | $1.29 | 2026-09-01 | |
| 2 | 96.4% | 54.7 | $0.103 | 2026-09-01 | |
| 3 | 96.2% | 65.3 | $1.15 | 2026-09-01 | |
| 4 | 95.6% | 60.6 | $0.785 | 2026-09-01 | |
| 5 | 95.4% | 59.5 | $0.401 | 2026-09-01 | |
| 6 | 95.4% | 58.8 | $0.338 | 2026-09-01 | |
| 7 | 95.0% | 65.1 | $2.05 | 2026-09-01 | |
| 8 | 93.4% | 60.7 | $0.760 | 2026-09-01 | |
| 9 | 93.0% | 56.2 | $0.043 | 2026-09-01 | |
| 10 | 92.0% | 52.4 | $0.018 | 2026-09-01 | |
| 11 | 88.8% | 57.2 | $0.0099 | 2026-09-01 | |
| 12 | 88.6% | 61.1 | $1.92 | 2026-09-01 | |
| 13 | 86.6% | 61.5 | $0.554 | 2026-09-01 | |
| 14 | 86.6% | 56.8 | $0.540 | 2026-09-01 | |
| 15 | 86.0% | 52.2 | $1.08 | 2026-09-01 | |
| 16 | 85.8% | 58.7 | $0.672 | 2026-09-01 | |
| 17 | 85.6% | 58.7 | $1.13 | 2026-09-01 | |
| 18 | 82.8% | 58.9 | $0.710 | 2026-09-01 | |
| 19 | 82.6% | 64.1 | $1.36 | 2026-09-01 | |
| 20 | 82.2% | 52.0 | $0.134 | 2026-09-01 | |
| 21 | 82.0% | 59.7 | $2.42 | 2026-09-01 | |
| 22 | 82.0% | 59.5 | $0.352 | 2026-09-01 | |
| 23 | 80.8% | 62.1 | $1.44 | 2026-09-01 | |
| 24 | 80.0% | 59.4 | $2.19 | 2026-09-01 | |
| 25 | 79.6% | 61.0 | $1.19 | 2026-09-01 | |
| 26 | 79.6% | 56.1 | $1.49 | 2026-09-01 | |
| 27 | 79.6% | – | – | 2026-09-01 | |
| 28 | 78.8% | 62.5 | $0.954 | 2026-09-01 | |
| 29 | 78.8% | 60.4 | $0.783 | 2026-09-01 | |
| 30 | 78.2% | 63.1 | $0.801 | 2026-09-01 | |
| 31 | 78.2% | 57.0 | $1.22 | 2026-09-01 | |
| 32 | 78.2% | 53.9 | $0.265 | 2026-09-01 | |
| 33 | 78.0% | 62.9 | $0.455 | 2026-09-01 | |
| 34 | 77.6% | 50.4 | $0.586 | 2026-09-01 | |
| 35 | 77.4% | 56.1 | $1.30 | 2026-09-01 | |
| 36 | 76.4% | 61.0 | $0.652 | 2026-09-01 | |
| 37 | 76.4% | 60.0 | $0.772 | 2026-09-01 | |
| 38 | 76.4% | 59.7 | $1.05 | 2026-09-01 | |
| 39 | 76.4% | 55.4 | $0.457 | 2026-09-01 | |
| 40 | 76.2% | 55.1 | $0.275 | 2026-09-01 | |
| 41 | 75.8% | 59.9 | $1.45 | 2026-09-01 | |
| 42 | 75.0% | 57.4 | $0.280 | 2026-09-01 | |
| 43 | 75.0% | 56.1 | $0.416 | 2026-09-01 | |
| 44 | 75.0% | 47.8 | $0.309 | 2026-09-01 | |
| 45 | 74.8% | 53.8 | $0.439 | 2026-09-01 | |
| 46 | 74.4% | 64.2 | $0.444 | 2026-09-01 | |
| 47 | 74.2% | 53.7 | $0.457 | 2026-09-01 | |
| 48 | 74.0% | 55.5 | $0.023 | 2026-09-01 | |
| 49 | 73.8% | 56.6 | $0.466 | 2026-09-01 | |
| 50 | 73.4% | 59.0 | $0.869 | 2026-09-01 | |
| 51 | 73.0% | 55.1 | $0.511 | 2026-09-01 | |
| 52 | 72.8% | 62.1 | $1.57 | 2026-09-01 | |
| 53 | 72.4% | – | $0.312 | 2026-09-01 | |
| 54 | 72.2% | 58.3 | $0.516 | 2026-09-01 | |
| 55 | 71.4% | 60.2 | $0.449 | 2026-09-01 | |
| 56 | 71.4% | 56.5 | $0.223 | 2026-09-01 | |
| 57 | 71.2% | 57.4 | $0.964 | 2026-09-01 | |
| 58 | 71.0% | 56.3 | $0.010 | 2026-09-01 | |
| 59 | 70.0% | 59.4 | $0.168 | 2026-09-01 | |
| 60 | 70.0% | 52.3 | $0.807 | 2026-09-01 |
AI coding benchmark leaderboard · All LLM benchmarks
Related benchmarks
Top head-to-heads on SWE-bench (Vals)
Cite as: BenchLeader, “SWE-bench (Vals) leaderboard”, data as of 15 Nov 2026, https://www.benchleader.com/benchmarks/vals_swebench.
SWE-bench (Vals): questions
- What does SWE-bench (Vals) 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 (Vals)?
- As of 15 Nov 2026, the highest score on SWE-bench (Vals) is 97.0% by Claude Opus 5 (result dated 1 Sep 2026), ahead of DeepSeek V4 Pro at 96.4%. Scores come from Vals AI and are re-read daily.
- Which AI model leads SWE-bench (Vals)?
- Claude Opus 5 leads SWE-bench (Vals) with 97.0% as of 15 Nov 2026 (result dated 1 Sep 2026), ahead of DeepSeek V4 Pro at 96.4%.
- How many models have SWE-bench (Vals) results?
- 86 model configurations have a SWE-bench (Vals) result on BenchLeader, all taken from Vals AI.
- Who runs SWE-bench (Vals) and how often is it updated?
- SWE-bench (Vals) is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does SWE-bench (Vals) count toward the BenchLeader Index?
- No. SWE-bench (Vals) is shown for reference but left out of the composite index.