Quadruped reinforcement learning task (Anthropic)
Claude Opus 5 31.3 leads since Jul 2026 Only in system cards
Agents train a quadruped to reach high performance in a continuous control task.
In-house Anthropic evaluation. The tasks are not public.
Made by Anthropic Released May 2025 Machine learning engineering Tasks kept private Last in a system card Jul 2026
Last model added Jul 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
- Agent task Many steps in a terminal, browser, desktop, app, game or robot.
8
Models scored in lab cards
12
System cards report it, from 1 lab
Claude Opus 5 leads with 31.3 among models released by Jul 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 Jul 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 15 more
- 1 Claude Opus 5 31.3
- 2 Claude Mythos Preview 30.9
- 3 Claude Opus 4.7 26.5
- 4 Claude Sonnet 5 19.9
- 5 Claude Sonnet 4.5 1.30
- 6 Claude Opus 4 1.25
- 7 Claude Opus 4.1 1.18
- 8 Claude Sonnet 4 0.35
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 31.3 | Jul 2026 | Model card | |
| 2 | Claude Mythos Preview | 30.9 | Apr 2026 | Model card | |
| 3 | Claude Opus 4.7 | 26.5 | Apr 2026 | Model card | |
| 4 | Claude Sonnet 5 | 19.9 | Jun 2026 | Model card | |
| 5 | Claude Sonnet 4.5 | 1.30 | Sep 2025 | Model card | |
| 6 | Claude Opus 4 | 1.25 | May 2025 | Model card | |
| 7 | Claude Opus 4.1 | 1.18 | Aug 2025 | Model card | |
| 8 | Claude Sonnet 4 | 0.35 | May 2025 | Model card |
4 models have held number one since its release in May 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Anthropic
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
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.
12 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.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 24 Jul 2026 | Claude Opus 5 system card | 31.3 | Quadruped RL, highest score, no hyperparameters variant (Table 2.3.5.A) | AI R&D risk |
| 30 Jun 2026 | Claude Sonnet 5 system card | 19.9 | Quadruped RL, highest score with no hyperparameters 19.94 (Table 2.3.3.A) | AI R&D risk |
| 16 Apr 2026 | Claude Opus 4.7 system card | 26.5 | Quadruped RL, highest score, no hparams variant (Table 2.3.4.A) | AI R&D risk |
| 7 Apr 2026 | Claude Mythos Preview system card | 30.9 | Quadruped RL, highest score, no hyperparameters variant, 30.87 (Table 2.3.3.A) | AI R&D risk |
| 17 Feb 2026 | Claude Sonnet 4.6 system card | Quadruped reinforcement learning, highest score 18.72 no hyperparameter variant and 17.88 no reward function variant, no overall (6.3.3.5) | AI R&D risk | |
| 5 Feb 2026 | Claude Opus 4.6 system card | Quadruped reinforcement learning, two variants with no overall: highest score 20.96 no hyperparameter variant and 21.99 no reward function variant, threshold 12 (8.3.3.5) | AI R&D risk | |
| 24 Nov 2025 | Claude Opus 4.5 system card | Quadruped reinforcement learning, easy variants with no overall: highest score 19.48 no hyperparameter variant and 19.16 no reward function variant, threshold 12 (7.3.2.5) | AI R&D risk | |
| 29 Sep 2025 | Claude Sonnet 4.5 system card | 1.3 | Quadruped reinforcement learning, easier variant, best run normalized score 1.302 (threshold 1), median below threshold (9.3.3.5) | AI R&D risk |
| 5 Aug 2025 | Claude Opus 4.1 system card addendum | 1.18 | Quadruped locomotion, easier variant, 1 in 30 trials above threshold, scoring 1.183 (6.4.1) | AI R&D risk |
| 22 May 2025 | Claude Sonnet 4 system card | 0.35 | Quadruped reinforcement learning, easy variants, highest score 0.35 (threshold 1). An earlier snapshot scored at threshold (7.3.3.5) | AI R&D risk |
| 22 May 2025 | Claude Opus 4 system card | 1.25 | Quadruped reinforcement learning, easy variant without hyperparameters, best run normalized score 1.25 (threshold 1) (7.3.3.5) | AI R&D risk |
| 24 Feb 2025 | Claude 3.7 Sonnet system card | Internal AI Research Evaluation Suite, continuous control task variants 1 and 2, score distribution chart without printed values. Text says very far from the human baseline of 1.0 (Figure 21) | AI R&D risk |
More 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%