GPT-5.2 Codex
gpt-5.2-codexgpt-52-codexopenai-gpt-52-codex- 🔧 Coding-specialized variant built on OpenAI's GPT-5.2 base model.
- 🧠 Tuned for long-horizon, agentic software engineering tasks.
- 📏 256K context, extended further via native context compaction.
- 🔒 Notably stronger defensive cybersecurity capabilities than prior Codex models.
- 🆕 Improved large refactors, code migrations, and Windows-environment reliability.
- 👁️ Supports reasoning, vision, function calling, and web search.
- ⚡ Offers low, medium, high, and xhigh reasoning-effort settings.
- 🏢 Geared toward professional, multi-hour repository-scale engineering work.
OpenAI is an American artificial intelligence research organization headquartered in San Francisco, structured as both a for-profit public benefit corporation and a nonprofit foundation. The lab developed the GPT family of large language models, the DALL-E image generation…
Explore 27 more models by OpenAI →GPT-5.2 Codex is OpenAI's coding-specialized model, an edition of GPT-5.2 further optimized for agentic software engineering inside the Codex platform. Released in 2025, it is built to handle long-horizon work: large-scale code transformations, refactors, migrations, and feature builds that span extended sessions. It supports four reasoning-effort levels (low, medium, high, and xhigh) and exposes reasoning, vision, function-calling, and web-search capabilities.
Compared with earlier Codex models such as GPT-5.1-Codex-Max, OpenAI reports improvements in long-context understanding, more reliable tool calling, improved factuality, and native compaction—the ability to compress history and keep working coherently past the base context window. This lets it operate in large repositories over extended sessions without losing track, even when plans change or attempts fail. OpenAI also highlights stronger performance in native Windows environments and significantly enhanced cybersecurity capabilities, calling it the strongest model on its internal Professional CTFs evaluation to date, aided by compaction on long-running cyber tasks.
OpenAI evaluates the model on SWE-Bench Pro, where it must generate a patch to solve realistic software engineering tasks, though specific figures should be confirmed against the official model card and addendum.
Within the Codex lineage, GPT-5.2 Codex was succeeded by GPT-5.3 Codex, which OpenAI describes as combining the Codex training stack with GPT-5's reasoning and general-intelligence stack.
This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.
| Seller | Reputation↓ | Routing | Input $/M | Cached $/M | Output $/M | Categories | API |
|---|---|---|---|---|---|---|---|
| Venice.ai Proxy 0x1f22…18c9 | 100.00 | #3 | $1.095 | $0.1095 | $8.75 | chat,reasoning,coding,vision,multimodal,web-search | openai-chat-completions |
| ▲ Apex Ant 0x73b4…e736 | 91.03 | #2 | $0.88 | $0.09 | $7.00 | chat,coding,tasks,vision,multimodal,reasoning,long-context,agents | openai-chat-completions |
| NovaRoute AI 0xc50d…ed7b | 67.69 | #1 | $0.21 | $0.21 | $1.68 | chat,coding,code,reasoning,tasks,developer,codex,gpt,surplus,openai-compatible,low-cost,verified,github,response-auth,base-usdc,monitored | openai-chat-completions |
| Fire Ant 🔥🐜 0xbe05…bc5d | 40.37 | gated | $1.2209 | $0.1221 | $9.7564 | — | — |
| Meridian AI 0x8c8c…06f5 | 5.64 | gated | $0.378 | $0.378 | $3.024 | chat,coding | openai-chat-completions |
| Apex TEE Test 0xe672…7955 | 0.51 | gated | $0.5913 | $0.0624 | $4.725 | chat,coding,tasks,vision,multimodal,reasoning,long-context,agents | openai-chat-completions |
| antseed-neon-puma-944e 0x6650…944e | 0.00 | gated | $0.7468 | $0.219 | $5.9675 | chat,coding,math | openai-chat-completions |
| Leftermute 0x388b…5389 | 0.00 | gated | $0.1991 | $0.1991 | $1.5908 | chat,coding,json,math,tools | openai-chat-completions |
"Best price" and the seller table are live AntSeed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = on-chain trust (0-100). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on AntSeed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.