OpenAIOpenAI·💬 Text Generation

GPT-5.3 Codex

ReasoningVisionCodeFunction CallingWeb Searchanonymized
🧠 Try in Intelligence →Try on Venice.ai ↗
Quick reference
GPT-5.3 Codex — TLDR
  • 🔧 OpenAI's Codex-native coding agent built on GPT-5.3.
  • 🆕 Unifies GPT-5.2-Codex coding with GPT-5.2 reasoning in one model.
  • ⚡ OpenAI reports it runs 25% faster than its predecessor.
  • 📏 Catalog lists a 400,000-token context window.
  • 👁️ Supports vision, function calling, and web search.
  • 🎯 Tunable reasoning effort: low, medium, high, xhigh.
  • 🧠 Targets long-horizon debugging, deployment, and software-lifecycle work.
  • 💬 Delivered through OpenAI's Responses API for agentic coding.
💰 Pricing
$2.19 / $17.50
per 1M · input / output
📏 Context
400K tokens
📅 On Venice since
Feb 24, 2026
99 days ago
Provider

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…

Read full profile →
24 models on Venice
13 text · 4 video · 2 image · 2 embedding · 2 inpaint · 1 asr
Since Jan 15, 2025

About this model

GPT-5.3 Codex is OpenAI's specialized coding model, described as a Codex-native agent that pairs frontier coding performance with general reasoning to support long-horizon, real-world technical work. It is the third major iteration of the Codex-specific line and the first to combine the Codex coding training stack with GPT-5.3's reasoning and general-intelligence base in a single unified model. Beyond writing and reviewing code, OpenAI positions it for the broader software lifecycle—debugging, deploying, monitoring, and writing product docs, tests, and metrics.

Compared with its same-family predecessor GPT-5.2 Codex, OpenAI states the new model advances both the frontier coding ability of GPT-5.2-Codex and the reasoning and professional-knowledge capabilities of GPT-5.2, together in one model that is also 25% faster. According to OpenAI, it sets new highs on SWE-Bench Pro and Terminal-Bench among the four benchmarks the company uses to measure coding, agentic, and real-world capability.

The model exposes selectable reasoning effort (low, medium, high, and xhigh) and is delivered through OpenAI's Responses API. A separate, smaller research-preview variant, Codex-Spark, targets near-instant iteration with a 128K context window and text-only input.

Within OpenAI's wider 2026 lineup, GPT-5.3 Codex sits alongside newer general-purpose releases such as GPT-5.4 and the later GPT-5.5.

This About section is AI-generated from public sources (Claude Opus 4.8), with no human editing. It may contain inaccuracies — verify critical details against the sources listed above.

Data sources: Venice API · HuggingFace · Wikipedia — enrichment updated 1d ago