February 5, 2026

OpenAI GPT-5.3-Codex released

OpenAIAn AI research company — created GPT, ChatGPT, DALL-E, and the o-series reasoning models. released GPTGenerative Pre-trained Transformer — OpenAI's family of large language models from GPT-1 through GPT-5.-5.3-Codex on February 5, 2026 — an agenticAI systems that plan and execute multi-step tasks autonomously — tools, browsing, and sub-agents. coding model merging Codex performance with GPTGenerative Pre-trained Transformer — OpenAI's family of large language models from GPT-1 through GPT-5.-5.2 reasoningStep-by-step logical thinking in AI models — chain-of-thought before answering hard problems., 25% faster in Codex tools.

What it was for

OpenAIGPTGenerative Pre-trained Transformer — OpenAI's family of large language models from GPT-1 through GPT-5.-5.3-Codex handled long-running research, tool use, and complex execution while staying steerable mid-task — OpenAIAn AI research company — created GPT, ChatGPT, DALL-E, and the o-series reasoning models.'s answer to Claude Opus 4.6 in the same week. OpenAIAn AI research company — created GPT, ChatGPT, DALL-E, and the o-series reasoning models. said early versions helped build the model itself, and it shipped in Codex app, CLICommand-line interface — interacting with software by typing text commands., and IDEIntegrated development environment — an editor with debugging, build tools, and project management built in. extensions on GB200 systems.

Companies

  • OpenAI

Why it's here

GPTGenerative Pre-trained Transformer — OpenAI's family of large language models from GPT-1 through GPT-5.-5.3-Codex unified OpenAIAn AI research company — created GPT, ChatGPT, DALL-E, and the o-series reasoning models.'s coding and reasoningStep-by-step logical thinking in AI models — chain-of-thought before answering hard problems. stacks in one agenticAI systems that plan and execute multi-step tasks autonomously — tools, browsing, and sub-agents. model.

Why it mattered

Interactive steering during long agent runs became a baseline Codex feature.

What it solved

Developers juggled separate Codex and GPTGenerative Pre-trained Transformer — OpenAI's family of large language models from GPT-1 through GPT-5. models for coding versus general reasoningStep-by-step logical thinking in AI models — chain-of-thought before answering hard problems..

Media

  • OpenAI
    ImageOpenAI

    OpenAI, Public domain, via Wikimedia Commons

Related