GPT-5.2 Codex
GPT‑5.2-Codex is a version of GPT‑5.2 further optimized for agentic coding in Codex, including improvements on long-horizon work through context compaction, stronger performance on large code changes like refactors and migrations, improved performance in Windows environments, and significantly stronger cybersecurity capabilities.
text + image + file → text · made by OpenAI
Sold by 9 ways
| Seller | Lane | Rate |
|---|---|---|
| Databricks cloud | standard | $1.75per Mtok in$14per Mtok out$0.17per Mtok cached$1.75per Mtok cache writeraw.githubusercontent.com · read 2026-09-16 |
| Microsoft Azure AI cloud | standard | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedmodels.dev · read 2026-09-16 |
| GitHub Copilot 0 ways | $1.75per Mtok in$14per Mtok out$0.17per Mtok cached | |
| GitHub Copilot aggregatornot on the seller's list today | standard | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedmodels.dev · read 2026-09-04 |
| Nous Research aggregator | standard | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedinference-api.nousresearch.com · read 2026-09-12 |
| OpenRouter aggregator | standard | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedopenrouter.ai · read 2026-08-24 |
| Requesty aggregator | azure | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Requesty aggregatornot on the seller's list today | openai-responses | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Vercel AI Gateway aggregator | standard | $1.75per Mtok in$14per Mtok out$0.17per Mtok cachedai-gateway.vercel.sh · read 2026-08-25 |
| Venice AI aggregator | standard | $2.19per Mtok in$17.5per Mtok out$0.22per Mtok cachedmodels.dev · read 2026-09-16 |
Measured 13 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 7thof 47 | SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent) | % Resolved | 72.8 |
| 10thof 25 | SWE-Bench Pro (Public Dataset) | Resolve Rate | 41.04 |
| 24thof 112 | Ale bench — Epoch AI | Performance | 1299.9 |
| 25thof 180 | SWE-bench Verified (Bash Only toggle off — all agents) | % Resolved | 72.8 |
| 27thof 42 | Design Arena · godotgamedev (agents) | Elo | 1142 |
| 28thof 65 | Apex agents — Epoch AI | Pass@1 score | 0.276 |
| 31stof 202 | Terminalbench — Epoch AI | Accuracy mean | 0.665 |
| 35thof 55 | Design Arena · androidnative (agents) | Elo | 1176 |
| 58thof 63 | Design Arena · mobileapps (agents) | Elo | 1117 |
| 59thof 62 | Design Arena · webapps (agents) | Elo | 1065 |
| 62ndof 64 | Design Arena · fullstack (agents) | Elo | 1002 |
| 98thof 121 | Webdev arena — Epoch AI | Arena Score | 1338.28 |
| 101stof 128 | LMArena · WebDev | Score (Elo) | 1339 |
About
GPT-5.2 Codex — a text model from OpenAI, sold by 7 companies from $1.75 in and $14 out per million tokens, placed 7th of 47 on SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent).
It takes text, images and files and returns text, with a context window of 400,000 tokens. It was published in December 2025, trained on material up to August 2025. Its sellers say it can reason step by step and call a tool. The catalogue files it under code and agents. Seven companies sell it. The cheapest is $1.75 in and $14 out per million tokens at Databricks. Beside the standard rate there is a separately routed lane. It has been measured on 13 boards, and stands best at 7th of 47 on SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent).
Every current figure
- Maker
- OpenAI
- Register
- model
- Takes
- text + image + file
- Returns
- text
- Context
- 400,000 tokens
- Longest answer
- 128,000 tokens
- Published
- December 2025
- Knowledge to
- August 2025
- Licence
- not read
- Sellers
- 7
- Maker's own price
- not read
- Price
- $1.75 in and $14 out per million tokens — Databricks
- Boards
- 13
- Best place
- 7th of 47 — SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent)
Known as 6 names
gpt-5.2-codexGPT 5.2 Codex5.2 codexopenai/gpt-5.2-codexazure/gpt-5.2-codex@eastus2openai-responses/gpt-5.2-codex