GPT-5.6-terra

A Hacker News user released a benchmark for flight search intent, testing AI models on parsing natural-language queries into structured search fields. GPT-5.6-sol led with an 87% pass rate (212/244), …

DeepSWE, a new benchmarking platform for AI coding agents, achieves clearer separation between frontier models than older benchmarks like SWE-bench by using contamination-free tasks across 91 reposito…