GPT-6 Astra vs MediaPipe on 3D hand pose: 3 min per frame vs 20 ms
chris_j_paxton · x · 2026-09-08
Developer kstonekuan benchmarked GPT-6 Astra on 3D hand pose estimation, reporting SOTA-level visual/spatial reasoning (MazeBench) and potential for labeling robotics datasets. Constrained to MediaPipe's output schema and visualized side by side, the model's high-reasoning mode took 3 minutes per frame versus MediaPipe's 20 ms — a 9000x gap. MediaPipe can't handle gloved hands, which is where a reasoning model could help. Commenter chrisjpaxton notes MediaPipe is old and most serious data teams have better in-house tools, but it remains highly useful.
Related event: GPT-6 Astra Shines in Hand Pose Estimation but Is 9000x Slower(2 posts)→
More from Embodied
- WanGP v18.81 adds Tencent's Hy-Nomas world model, playable on 8GB VRAM — cocktailpeanut · 2026-10-27
- NVIDIA open-sources FLORA, a robotics data engine ~125x cheaper than GPT-6 Astra — nvidia · 2026-10-27
- Embodied AI startups need a different operational discipline, investor says — ___Mufasaa · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27