GPT-6 Astra official benchmarks: 98.6% on ARC-AGI-3
OpenAI released GPT-6 Astra, which Greg Brockman called the "AGI era," with ARC-AGI-3 jumping from 7.8% in the previous generation to 98.6%, alongside other benchmark scores and a phased rollout plan.
OpenAI launched GPT-6 Astra. Greg Brockman's words: "Welcome to the AGI era." Here are the benchmarks. → ARC-AGI-3: 98.6%. GPT-5.6 Sol scored 7.8%. → FrontierMath Tier 4: 97.6%. Sol 83%. Fable 5.1 87.8%. → Terminal-Bench Science: 64.6%. Sol 22.4%. Fable 5.1 52.6%. → AutomationBench: 41.4%. Sol 18.1%. Fable 5.1 31.4%. → ExploitBench: 100%. Sol 78.5%. → SRE-Bench: 99.2%. Sol 68.7%. → Auto-review circumvention: 0%. ARC-AGI was designed to be the one benchmark AI couldn't brute-force. It measures something simple: can the AI think through a problem it's never seen before? Actually reason. Sol scored 7.8%. Astra scored 98.6%. Rolling out now. Limited organizations first. ChatGPT Plus, Pro, Business, and Enterprise users over the coming days. API and AWS after that.
This post quotes another post: “GPT-6 Astra announced with a claim to do anything on a computer”