Last month, OpenAI launched GPT-5.1-Codex-Max, a new frontier model designed for long-running, agentic coding tasks. This model was trained to operate across multiple context windows involving more than a million tokens through a technique called compaction. In internal testing, OpenAI claims that GPT‑5.1-Codex-Max was able to work on tasks for more than 24 hours.
The regular GPT-5.1-Codex model scored 73.7% on SWE-Bench Verified, 66.3% on SWE-Lancer IC SWE, and 52.8% on TerminalBench 2.0. However, GPT-5.1-Codex-Max delivered even higher results, achieving 77.9% on SWE-Bench Verified, 79.9% on SWE-Lancer IC SWE, and 58.1% on TerminalBench 2.0.
This new developer-focused model is now available in the API with the same pricing ($1.25 / 1M input tokens and $10 / 1M output tokens) and rate limits as GPT-5. Previously, it was accessible only through the Codex CLI, IDE extension, cloud, and code review for users with ChatGPT Plus, Pro, Business, Edu, and Enterprise subscriptions. Since this model is faster and more token-efficient than the regular GPT-5.1-Codex, OpenAI is making GPT-5.1-Codex-Max the default model in Codex.
While the previous Codex models were mostly optimized to run on Unix-based environments, the GPT-5.1-Codex-Max model is trained to operate in Windows environments. So, if you are a developer using a Windows environment, switching to this new model is a no-brainer.
Thanks to the API access, this new model is now available inside various coding tools, including Cursor, GitHub Copilot, Linear, and others. This new model will be available to GitHub Copilot Pro, Pro+, Business, and Enterprise subscription users, and it will be available in the model picker in the following locations:
- Visual Studio Code in all modes: chat, ask, edit, agent
- Copilot Chat in github.com
- Copilot Chat in GitHub Mobile (iOS and Android)
- Copilot CLI
OpenAI’s main competitor, Anthropic, is now focused on developing the best coding models. Their focus has paid off well since their Claude Code tool has become the fastest-growing SaaS product to reach the billion-dollar milestone. In fact, Anthropic may hit $8-$10 billion in revenue this year by mostly selling coding-focused models to developers and enterprises.
To better compete with Anthropic, OpenAI released GPT-5-Codex in August. Codex usage has grown more than 10 times since the launch of GPT-5-Codex in August, with Codex models now serving trillions of tokens per week. It remains to be seen if this new Codex model release will be enough to slow Anthropic’s remarkable momentum in the enterprise coding market.