OpenAI's GPT-5.1-Codex-Max
OpenAI has released their latest model GPT‑5.1-Codex-Max which is focused on agentic coding.
GPT‑5.1-Codex-Max is built for long-running, detailed work. It’s our first model natively trained to operate across multiple context windows through a process called compaction, coherently working over millions of tokens in a single task. This unlocks project-scale refactors, deep debugging sessions, and multi-hour agent loops.
From the benchmark that they’ve shared and whatever tweet that I’ve seen on Twitter/X, it looks like a good model.
Right now it is available in Codex and it should be available in API soon.
