GPT-5.3-Codex

2026-02-05 · Publisher : OpenAI · Domain : Artificial Intelligence

OpenAI launches GPT-5.3-Codex, its most capable agentic coding model, which combines the performance of GPT-5.2-Codex with the reasoning of GPT-5.2 and runs 25% faster.

What changes

  • First model described by OpenAI as having contributed to its own creation: the Codex team used it to debug its training and manage its deployment.
  • Extends the Codex agent beyond code: documents, spreadsheets, data analysis, with interaction during a task without loss of context.
  • First model rated "High" in cybersecurity in the Preparedness Framework, with the launch of the Trusted Access for Cyber pilot program.
  • Trained and served on NVIDIA GB200 NVL72 systems.

### Specifications - Context: 400,000 tokens, maximum output 128,000 tokens, text and image input, text output. - Reasoning effort: low, medium, high, xhigh. Knowledge cutoff: August 31, 2025. - Availability: Codex (app, CLI, IDE extension, web) with a paid ChatGPT subscription from February 5; Responses API on February 24. - API identifier: gpt-5.3-codex. ### Pricing | Mode | Input | Cached input | Output | | --- | --- | --- | --- | | Standard | $1.75 | $0.175 | $14.00 | | Fast mode | $3.50 | $0.35 | $28.00 | Price per million tokens (USD), OpenAI pricing page. ### Benchmarks | Evaluation | GPT-5.3-Codex | GPT-5.2-Codex | GPT-5.2 | | --- | --- | --- | --- | | SWE-Bench Pro (public) | 56.8% | 56.4% | 55.6% | | Terminal-Bench 2.0 | 77.3% | 64.0% | 62.2% | | OSWorld-Verified | 64.7% | 38.2% | 37.9% | | GDPval (wins or ties) | 70.9% | n/a | 70.9% | | Cybersecurity CTF | 77.6% | 67.4% | 67.7% | | SWE-Lancer IC Diamond | 81.4% | 76.0% | 74.6% | Scores published by the vendor. All at xhigh effort. ### Key takeaways GPT-5.3-Codex turns Codex from a coding assistant into a general-purpose agent on the workstation. It was later absorbed by GPT-5.4, which inherits its coding capabilities.

Links

Tags : LLM, OpenAI, GPT-5, Codex, Agent de code