Build more with GPT-5.1-Codex-Max

OpenAI has launched GPT-5.1-Codex-Max, a frontier agentic coding model, now available in Codex's CLI, IDE extension, cloud, and code review, with API access coming soon. The model is built on an update to the reasoning model and is OpenAI's first model trained natively with compaction across multiple context windows, allowing it to coherently process millions of tokens in a single task; in internal evaluations it has worked for over 24 hours continuously. On SWE-bench Verified, it performs better than GPT-5.1-Codex at medium reasoning effort, with 30% fewer thinking tokens, and adds an xhigh reasoning effort. On safety, Codex runs in a sandbox by default, with file writes limited to the workspace and network access off by default; enabling internet or web search introduces prompt injection risk. Under its Preparedness Framework, the model's cybersecurity capability has not reached High, but it is the strongest cybersecurity model it has deployed.