Leo Linsky Claims GPT-6-Astra Leads Multi-Agent Tests

Leo Linsky posted that his evaluation of GPT-6-Astra covered 100 complex multi-agent coding environments with competing and cooperating models in open-ended tasks. He stated the model is the new frontier leader by a wide margin and more dominant than the Fable 5 release. The post notes it outperforms the second-best model and includes an attachment showing the GBENCH Intelligence Benchmark leaderboard table.