← All models

GPT-5.1-Codex-Max

Published Publisher OpenAIgpt-5.1-codex-max

Overview

GPT-5.1-Codex-Max is an agentic coding model optimized for long-running software engineering tasks. OpenAI announced it on November 19, 2025, initially in Codex; it reached the Responses API on December 4, 2025. (openai.com)

Capabilities and uses

The model was trained for tasks including pull-request creation, code review, frontend development and coding questions. Its context-compaction capability lets it work across multiple context windows, supporting project-scale refactoring, deep debugging and multi-hour agent workflows. OpenAI reported a 77.9% score on SWE-bench Verified using xhigh reasoning effort. (openai.com)

Specifications and limitations

  • Context window: 400,000 tokens; maximum output: 128,000 tokens.
  • Modalities: text input and output; image input only. Audio and video are not supported.
  • API: supported through the Responses API, not Chat Completions. Fine-tuning is not supported.
  • Intended use: OpenAI recommended the model for agentic coding in Codex or similar environments, rather than general-purpose chat. Agentic workflows can be exposed to prompt injection and require appropriate safeguards and review. (developers.openai.com)

Availability

The API model was shut down on July 23, 2026. The original announcement's Codex availability describes its launch, not necessarily present-day access. (developers.openai.com)

API pricing

The official model page lists the following USD per 1 million text tokens: input $1.25, cached input $0.125, and output $10.00. These are listed API token prices, not a Codex subscription price; the model is now shut down. Source: OpenAI model and pricing page. Checked November 17, 2026. (developers.openai.com)

Official sources