π0.7
Physical Intelligence·
Roboticsroboticscloud
Overview
Steerable robot foundation model from Physical Intelligence, presented as the point where VLAs start generalising rather than replaying training data. Physical Intelligence reports 82.1% success on trained tasks and 47.3% zero-shot on unseen ones across seven robot embodiments and 50+ manipulation tasks, and demonstrates the model operating an air fryer it had only encountered in two fragmentary training episodes. Weights are not public.
Capabilities and innovations
Vision-language-action robot controlZero-shot spatial generalizationSemantic steering from language aloneMulti-step compositional manipulationEmergent task generalization beyond the training distributionOne policy across seven robot embodiments
Benchmarks
No tracked academic benchmark applies to robotics models; the timeline carries no scores and no leadership tracking for them.
- GPQA Diamond: not reported by the vendor
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: not reported by the vendor
- MMLU: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: not reported by the vendor
- MMMU-Pro: not reported by the vendor
Links
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.