OpenAI launches GPT-5.1-Codex-Max for long-horizon coding

What happened

OpenAI released GPT-5.1-Codex-Max, a new coding model built for extended software engineering workflows. The model uses a “compaction” technique to handle context across millions of tokens, enabling multi-hour coding, debugging, and refactoring sessions. It replaces GPT-5.1-Codex as the default for coding tasks and is now available to ChatGPT Plus, Pro, Business, Edu, and Enterprise users.

Why it matters

• The model improves token efficiency, using about 30% fewer reasoning tokens than its predecessor on coding benchmarks.
• It supports multi-file projects and extended coding loops, making it suitable for complex, long-horizon development tasks.
• The release enhances compatibility with Windows environments and CLI workflows.

Key facts

1. 1. GPT-5.1-Codex-Max is optimized for long-running coding tasks and uses a “compaction” technique for context management.
2. 2. The model is about 30% more token-efficient than GPT-5.1-Codex on SWE-Bench Verified benchmarks.
3. 3. It is available immediately for ChatGPT Plus, Pro, Business, Edu, and Enterprise users.
4. 4. API access is not yet available but is expected soon.
5. 5. The model supports multi-file projects and extended coding sessions.

Key Voices

• Not stated in the article.