GPT-6 Astra just destroyed the robot control benchmark with 95% accuracy vs Fable 5.1's 40%.
But here's the real flex: 6.2x fewer tokens and 2.3x cheaper.
This isn't just better performance, it's better economics. AI agents just got way more viable at scale.
But here's the real flex: 6.2x fewer tokens and 2.3x cheaper.
This isn't just better performance, it's better economics. AI agents just got way more viable at scale.