GPT-5.1 Codex Max

According to a Digital Applied migration guide, OpenAI scheduled 13 model snapshots for shutdown on July 23, 2026, including every pre-GPT-5.3 Codex variant — explicitly listing gpt-5-codex, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini, and gpt-5.2-codex among the five retiring Codex-lineage coding models. The The same guide warns that the replacements are not a one-to-one swap: configs, CI scripts, or agent backends that still pin the gpt-5.1-codex-max snapshot ID will fail after the cutoff and require active testing against successor Codex variants rather than a simple find-and-replace migration. A further wave of 12 or mo

Editor’s note (April 15, 2026): We updated this post to include information about the deprecation of GPT 5.1. We have deprecated the following models across all GitHub Copilot experiences (including Copilot Chat, inline edits, ask and agent modes, and code completions) on April 1, 2026. | Model | Deprecation date | Sug

The release of this new model comes shortly after a competing agent-based coding model, Claude Opus 4.6, was introduced by rival AI startup Anthropic.

GPT-5.1-Codex-Max is generally available to Copilot Enterprise, Copilot Business, Copilot Pro, and Copilot Pro+.

GPT-5.1-Codex-Max is now rolling out in public preview in GitHub Copilot. Availability in GitHub Copilot. OpenAI's GPT-5.1-Codex-Max will be...

OpenAI is updating its coding environment with GPT-5.1-Codex-Max, a new model designed to handle massive amounts of context and solve complex engineering tasks that take hours to complete.

OpenAI Debuts GPT-5.1-Codex-Max, a Long-Horizon Agentic Coding Model With Compaction for Multi-Window Workflows.

BenchGecko's third-party listing confirms GPT-5.1-Codex-Max as OpenAI's agentic coding model for long-running, high-context software work, with a 400K context window, multimodal support, and $1.25/$10 per million token pricing. It averages 85.6 across benchmarks, ranked 13th, with 72.0 average across 8 LiveBench tests. Top category scores are reasoning 84.6%, math 83.7%, and coding 81.4%. Additional LiveBench results include agentic coding at 56.7 and data analysis at 54.9. The page lists GPT-5.1-Codex-Max within the OpenAI GPT-5.1 family alongside GPT-5.1, GPT-5.1 Chat, GPT-5.1-Codex, and GPT-5.1-Codex-Mini, showing it scoring 3.4 points above GPT-5.1-Codex at the same $1.25/M input price and 400K context.

OpenAI published the GPT-5.1-Codex-Max System Card on November 18, 2025, documenting baseline safety evaluations covering disallowed content, jailbreaks, and vision capabilities. Product-specific mitigations include an agent sandbox and controlled network access, while model-specific mitigations address harmful tasks, prompt injection, and avoidance of data-destructive actions through safety training. The card also covers Preparedness capabilities assessment, including biological and chemical risk categories such as long-form biological risk questions, multimodal troubleshooting virology, ProtocolQA open-ended, tacit knowledge, and TroubleshootingBench. Only the first three of twenty-seven pages were retrieved, so capability detail beyond these sections is unavailable.

The OpenAI API reference documents GPT-5.1-Codex-Max as a version of GPT-5.1-codex optimized for long-running tasks, available only in the Responses API. It supports a 400,000-token context, 128,000 max output tokens, text and image input with text output, and a September 30, 2024 knowledge cutoff. Pricing is $1.25 per 1M input tokens, $0.125 cached input, and $10.00 per 1M output tokens. The page lists supported features including reasoning effort levels, streaming, and tool use, and notes a prompting guide for getting the most from the model. Endpoints such as Chat Completions, Realtime, Assistants, Fine-tuning, and Embeddings are not used; only Responses and Batch support this model.

OpenAI introduced GPT-5.1-Codex-Max on November 19, 2025 as a frontier agentic coding model available in Codex (CLI, IDE extension, cloud, code review), with API access noted as coming soon. It is the first OpenAI model natively trained to operate across multiple context windows via compaction, enabling project-scale refactors and multi-hour agent loops over millions of tokens. It is also the first trained for Windows environments and improved Codex CLI collaboration. Benchmarks show GPT-5.1-Codex-Max scoring 79.9% on SWE-Lancer IC versus 66.3% for GPT-5.1-Codex (high), and it beats GPT-5.1-Codex on SWE-bench Verified while using 30% fewer thinking tokens at medium reasoning effort. A new Extra High reasoning effort is introduced for non-latency-sensitive work. All evals were run with compaction enabled at Extra High reasoning effort.

The ChatGPT & Codex changelog documents ongoing Codex CLI releases, including version 0.155.1 in September 2026, which fixed a bug where new local TUI sessions left reasoning summaries enabled by default; providers that do not support reasoning summaries had been rejecting such requests. Explicit reasoning-summary sett The same changelog also tracks ChatGPT desktop app updates across 2026, with a September 2026 iOS entry adding folder support from the file picker, writing blocks with draft alternatives, and various worktree, queued prompt, and voice-call reliability fixes. It does not, however, contain any explicit entry naming GPT-5