Benchmark Almanac

SWE-bench Verified (hard subset)

Claude Sonnet 4.6 48.2% leads since Feb 2026 Only in system cards

Agents solve a subset of SWE-bench Verified tasks estimated to take a human engineer more than one hour.

Anthropic RSP autonomy evaluation built on public SWE-bench Verified tasks.

Made by Anthropic Released Feb 2025 Machine learning engineering Tasks kept private Last in a system card Feb 2026 Last model added Feb 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Code A program or patch run against tests.
9
Models scored in lab cards
9
System cards report it, from 1 lab

Claude Sonnet 4.6 leads with 48.2% among models released by Feb 2026

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

The newest model with a score here came out in Feb 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 48.2 Claude Sonnet 4.6
  2. 47.2 Claude Opus 4.6
  3. 46.7 Claude Opus 4.5
  4. 45.3 Claude Sonnet 4.5
  5. 43.8 Claude Opus 4.1
  6. 39.5 Claude Opus 4
  7. 36.7 Claude Sonnet 4
  8. 36.6 Claude Haiku 4.5
  9. 23.0 Claude 3.7 Sonnet
  1. 1 Claude Sonnet 4.6 48.2%
  2. 2 Claude Opus 4.6 47.2%
  3. 3 Claude Opus 4.5 46.7%
  4. 4 Claude Sonnet 4.5 45.3%
  5. 5 Claude Opus 4.1 43.8%
  6. 6 Claude Opus 4 39.5%
  7. 7 Claude Sonnet 4 36.7%
  8. 8 Claude Haiku 4.5 36.6%
  9. 9 Claude 3.7 Sonnet 23%
#ModelScoreSettingDateSource
1 Claude Sonnet 4.6 48.2% Feb 2026 Model card
2 Claude Opus 4.6 47.2% Feb 2026 Model card
3 Claude Opus 4.5 46.7% Nov 2025 Model card
4 Claude Sonnet 4.5 45.3% Sep 2025 Model card
5 Claude Opus 4.1 43.8% Aug 2025 Model card
6 Claude Opus 4 39.5% May 2025 Model card
7 Claude Sonnet 4 36.7% May 2025 Model card
8 Claude Haiku 4.5 36.6% Oct 2025 Model card
9 Claude 3.7 Sonnet 23% Feb 2025 Model card

7 models have held number one since its release in Feb 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Lab claims Not tested since Feb 2026
Claude Sonnet 4.6 48.2%

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Sonnet 4.6 Claude Opus 4.5 Claude Opus 4.1 Claude Opus 4 Claude 3.7 Sonnet 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Sonnet 4.6 Claude Opus 4

Labs report it only in system cards

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.

9 system cards report it, from 1 lab

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
17 Feb 2026 Claude Sonnet 4.6 system card 48.2 SWE-bench Verified (hard subset), 21.7/45 problems solved on average over 10 trials (6.3.2) AI R&D risk
5 Feb 2026 Claude Opus 4.6 system card 47.2 SWE-bench Verified (hard subset), 21.24/45 problems solved on average over 10 trials (8.3.2) AI R&D risk
24 Nov 2025 Claude Opus 4.5 system card 46.7 SWE-bench Verified (hard subset), 21/45 problems solved, averaged over 10 trials (7.3.1) AI R&D risk
15 Oct 2025 Claude Haiku 4.5 system card 36.6 SWE-bench Verified hard subset, 16.45/45 problems solved on average, pass@1 (6.3) AI R&D risk
29 Sep 2025 Claude Sonnet 4.5 system card 45.3 SWE-bench Verified (hard subset), 20.4/45 problems passed on average over 10 runs (9.3.2) AI R&D risk
5 Aug 2025 Claude Opus 4.1 system card addendum 43.8 SWE-bench Verified (hard subset), 18.4/42 problems solved on average, pass@1 (6.4.1) AI R&D risk
22 May 2025 Claude Sonnet 4 system card 36.7 SWE-bench Verified (hard subset), 15.4/42 problems passed on average over 10 runs (7.3.1) AI R&D risk
22 May 2025 Claude Opus 4 system card 39.5 SWE-bench Verified (hard subset), 16.6/42 problems passed on average over 10 runs (7.3.1) AI R&D risk
24 Feb 2025 Claude 3.7 Sonnet system card 23 SWE-bench Verified (hard subset), average of 9.65/42 problems passed, stated as a 23% success rate (7.2, 7.2.2) AI R&D risk

More in AI research

All 83 in AI research
  • MLS-Bench In use Agent must invent one algorithmic change in an ML research scaffold whose gain transfers across settings. Scored relative to human baselines. Claude Fable 5.1 50.3%
  • PaperBench In use Agent replicates ICML 2024 papers from scratch. LLM judge grades against author-written hierarchical rubrics. Qwen 3.8 Max 93%
  • MLE-bench In use Kaggle-style ML engineering competitions. Agent trains models and is scored against competition leaderboards. GPT-6 Astra 93.8%
  • PostTrainBench 1.2 In use CLI agent gets a base model, one H100 and a time budget to post-train it for a target benchmark. Scored on normalized gain. Claude Fable 5.1 44.6%
  • KernelBench In use Replace PyTorch operators with faster custom GPU kernels. Scored on correctness and speedup. MiniMax-M3 28.8%
  • OpenAI PRs Only in system cards Agents replicate internal OpenAI pull requests from a pre-PR branch and a prompt. Graded by hidden unit tests as pass@1. GPT-5.2 55%