Version of GPT-5.1 optimized for agentic coding in Codex.
GPT-5.1 Codex is a variant of GPT-5.1 purpose-built for agentic coding in Codex — OpenAI's agentic coding environment spanning CLI, IDE extension, cloud sandbox, and GitHub integration. Announced alongside GPT-5.1 on November 12, 2025, it is the direct successor to GPT-5 Codex and was itself superseded by GPT-5.1-Codex-Max (November 18, 2025) as the default coding model in Codex surfaces. GPT-5.1 Codex was trained on real-world software engineering tasks: building full projects from scratch, adding features and tests, debugging, large-scale refactors, and code reviews.
From a TheRouter operator's perspective, GPT-5.1 Codex is accessible through the standard /v1/chat/completions endpoint with full OpenAI-compatible parameters (temperature, top_p, max_tokens, tools, tool_choice, response_format, reasoning_effort, stop). It shares GPT-5.1's 400K context window, 128K max output, and $1.25/M input / $10/M output pricing. The model is listed as listed:true and active:true in TheRouter's catalog. For agentic coding workflows specifically, GPT-5.1 Codex benefits from dedicated Codex toolchain support (apply_patch, shell) first introduced with GPT-5.1.
Position within the family: GPT-5.1 Codex sits between the general-purpose GPT-5.1 (best for general reasoning/coding) and GPT-5.1-Codex-Max (best for long-running tasks with compaction). The Codex family also includes GPT-5.1-Codex-Mini (faster, cheaper) and GPT-5.1-Codex-Max (the most capable, with cross-window compaction). As of May 2026, GPT-5.1 Codex has been superseded by GPT-5.1-Codex-Max as the default Codex model, but remains available via API for agents that need the original agentic coding profile without the compaction overhead of Codex-Max.
| Type | Rate |
|---|---|
| Input | $1.35 / 1M tokens |
| Output | $10.80 / 1M tokens |
| Release date | 2025-11-12openai.com ↗ | verified |
| OpenAI API model id | gpt-5.1-codexopenai.com ↗ | verified |
| Model type | Agentic coding variant of GPT-5.1 — optimized for Codex and Codex-like harnesses with apply_patch and shell tool supportopenai.com ↗ | verified |
| Reasoning type | Adaptive — dynamically adjusts thinking depth; reasoning_effort supports 'none', 'low', 'medium', 'high'openai.com ↗ | verified |
| Dedicated tooling | apply_patch (freeform diff-based file editing), shell (command execution in sandboxed environment)openai.com ↗ | verified |
| Successor model (as of May 2026) | GPT-5.1-Codex-Max (released 2025-11-18; SWE-bench 77.9%, compaction, 24h+ runs)openai.com ↗ | verified |
| License | Proprietary (OpenAI API terms of service) | verified |
| Benchmark | Distribution | Score | Source |
|---|---|---|---|
SWE-bench Verified GPT-5.1 Codex at 'high' reasoning effort. Source: OpenAI GPT-5.1-Codex-Max appendix (comparison table). By comparison, GPT-5.1 general-purpose variant scores 76.3% on the same benchmark. | 73.7%% | openai.com ↗ | |
SWE-Lancer IC SWE IC (individual contributor) category. Source: OpenAI GPT-5.1-Codex-Max appendix. GPT-5.1-Codex-Max scores 79.9% on the same eval. | 66.3%% | openai.com ↗ | |
Terminal-Bench 2.0 Source: OpenAI GPT-5.1-Codex-Max appendix. Measures ability to complete software engineering tasks through terminal interactions. | 52.8%% | openai.com ↗ | |
Token efficiency — fast 10% On the fastest 10% of Codex tasks (by model-generated tokens), GPT-5.1 Codex uses 93.7% fewer tokens than GPT-5. On the slowest 10%, it spends 2× more tokens reasoning and editing. Source: OpenAI Codex upgrades announcement. | 93.7% fewer tokens than GPT-5 | openai.com ↗ |
Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.
curl https://api.therouter.ai/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $THE_ROUTER_API_KEY" -d '{
"model": "openai/gpt-5.1-codex",
"messages": [
{"role": "user", "content": "Summarize the key points from this input."}
]
}'GPT-5.1 Codex via TheRouter uses the same OpenAI-compatible endpoint. Just switch the base URL and model name.
curl https://api.therouter.ai/v1/chat/completions \
-H "Authorization: Bearer $THEROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-5.1-codex",
"messages": [
{"role": "user", "content": "Refactor the authentication module to use async/await instead of callbacks"}
],
"tools": [
{"type": "function", "function": {"name": "apply_patch", "parameters": {"type": "object", "properties": {"file_path": {"type": "string"}, "diff": {"type": "string"}}}},
{"type": "function", "function": {"name": "shell", "parameters": {"type": "object", "properties": {"command": {"type": "string"}}}}
],
"reasoning_effort": "medium"
}'OpenAI released GPT-5.1-Codex-Max, a new frontier agentic coding model with native compaction for multi-window context across millions of tokens. Achieves 77.9% on SWE-bench Verified and 79.9% on SWE-Lancer IC SWE. Replaces GPT-5.1 Codex as the default in all Codex surfaces. Available in CLI, IDE extension, cloud, and code review.
OpenAI released GPT-5.1 with adaptive reasoning (dynamically adjusts thinking depth) and introduced GPT-5.1 Codex optimized for agentic coding. New tools: apply_patch (freeform diff-based editing) and shell. Extended prompt caching to 24 hours. Pricing unchanged from GPT-5.
GPT-5.1 is the general-purpose reasoning model for all text-based and coding tasks. GPT-5.1 Codex is a specialized variant trained specifically for agentic coding in Codex — it excels at multi-file refactors, code review, automated patch generation, and long-running agent loops. For general code-chat or one-shot code generation, GPT-5.1 works just as well at the same price. For agentic coding with apply_patch/shell tools, use GPT-5.1 Codex.
No — GPT-5.1 Codex remains available via the OpenAI API and through TheRouter. It has been superseded by GPT-5.1-Codex-Max as the default model in Codex surfaces (CLI, IDE, cloud), but the original GPT-5.1 Codex is still an active, listed model in TheRouter's catalog. Route to GPT-5.1-Codex-Max if you need compaction for very long contexts. Route to GPT-5.1 Codex if you want the original agentic coding profile without compaction overhead.
Yes — GPT-5.1 Codex is accessible through the standard OpenAI /v1/chat/completions endpoint via TheRouter. It supports all standard OpenAI parameters. However, the dedicated apply_patch and shell tools are designed for the Responses API and Codex-like harnesses. For simple code generation or chat, the general GPT-5.1 model is recommended instead.
OpenAI recommends 'none' for latency-sensitive workloads (fastest responses), 'low' or 'medium' for standard coding tasks, and 'high' when reliability and correctness matter most. The model defaults to 'none'. GPT-5.1 Codex with 'medium' effort achieves strong coding performance efficiently, and even at 'high' effort it dynamically reduces thinking on simpler tasks.
| source | URL | retrieved | |
|---|---|---|---|
| Release date | openai.com ↗ | 2026-05-28 | verified |
| OpenAI API model id | openai.com ↗ | 2026-05-28 | verified |
| Model type | openai.com ↗ | 2026-05-28 | verified |
| Reasoning type | openai.com ↗ | 2026-05-28 | verified |
| Dedicated tooling | openai.com ↗ | 2026-05-28 | verified |
| Successor model (as of May 2026) | openai.com ↗ | 2026-05-28 | verified |
| License | — | — | verified |
| SWE-bench Verified | openai.com ↗ | 2026-05-28 | verified |
| SWE-Lancer IC SWE | openai.com ↗ | 2026-05-28 | verified |
| Terminal-Bench 2.0 | openai.com ↗ | 2026-05-28 | verified |
| Token efficiency — fast 10% | openai.com ↗ | 2026-05-28 | verified |
| GPT-5.1-Codex-Max released — supersedes GPT-5.1 Codex as default in Codex | openai.com ↗ | 2026-05-28 | verified |
| GPT-5.1 and GPT-5.1 Codex announced for developers | openai.com ↗ | 2026-05-28 | verified |
| What is the difference between GPT-5.1 and GPT-5.1 Codex? | openai.com ↗ | 2026-05-28 | to verify |
| Is GPT-5.1 Codex deprecated? | openai.com ↗ | 2026-05-28 | to verify |
| Can I use GPT-5.1 Codex outside of Codex? | openai.com ↗ | 2026-05-28 | to verify |
| What reasoning_effort should I use with GPT-5.1 Codex? | openai.com ↗ | 2026-05-28 | to verify |