GPT-6 Astra represents a step-function change in model capability for interactive reasoning problems. It scores 66% on ARC-AGI-3 using our standard harness, and nearly 100% with a continuous conversation harness and custom compaction, at a cost of roughly $360 per game. In fact,…↗

GPT-6-Astra Benchmarks ARC-AGI-3 - 98.6% FrontierMath Tier 4 v2 - 97.6% DeepSWE v1.1 - 74.1% ExploitBench - 100%↗

GPT-5.6 Sol has been used to solve open problems in mathematics. So why was it struggling with ARC-AGI-3, a benchmark of 2D puzzle games? We investigated. The harness was not letting it remember what it had learned. We found that enabling two API settings tripled our scores wi…↗

lol did nobody at Anthropic stop for a second and wonder why the numbers looked this absurd before posting the “victory”-tweet? openai.com/index/how-two-…↗