OpenAI's Boris Power: 80-90% of research targets GPT-7/8, GPT-6 crushes robotics benchmarks
ccerrato147 · x · 2026-09-27
OpenAI Head of Applied Research Boris Power told Fireside Alpha that 80-90% of the company's research targets GPT-7, GPT-8 and beyond, as most fundamental value lies in future models. He outlined a distill pipeline: train a giant GPT-6 Astra teacher, then distill it into a small GPT-Luna model for specialized use. On robotics, he said GPT-6 is already crushing benchmarks, training data has long been the field's biggest bottleneck, robot bodies are now "good enough," and if the brain is the only bottleneck, robot parts could face a shortage akin to the GPU crunch. The brief also notes code-mined evidence of a higher-tier ChatGPT Pro Max plan and OpenAI researchers praising Cerebras-powered models' speed.
Related event: OpenAI Says 80-90% of Research Aims at GPT-7/8(2 posts)→
More from Embodied
- Tesla Optimus scales to hundreds per week but hands remain a bottleneck — AIFlow_ML · 2026-09-27
- $800 litter-picking robot MOSS hits V0.3, open-source V0.4 already printing — AIFlow_ML · 2026-09-27
- GPT-6 Astra Robot Demos Look Great on X, But Real-Time Footage Shows Painful Slowness — DJiafei · 2026-09-27
- VLA-REPLICA benchmark accepted at NeurIPS 2026, adds MolmoAct 2 and GR00T N1.7 — YuXiang_IRVL · 2026-09-27
- Humanoid Robotics Debate: Claims US Firms Fake Demos While Chinese Companies Stay Legit — cixliv · 2026-09-27
- Robots crack whips with zero learning: motion capture retargeting enables dynamic manipulation — ATTlKA · 2026-09-27