Meet GPT‑5.3 Codex, OpenAI’s most advanced agentic coding model for building autonomous, tool‑using developer agents and production‑grade AI workflows. Unlike a simple code autocomplete tool, GPT‑5.3‑Codex can initiate and sustain long-running tasks: conducting research, using external tools, managing deployments, writing documentation, and iterating on complex software over millions of tokens — all while keeping you actively in the loop.
Achieves state-of-the-art performance on SWE-Bench Pro — a rigorous, multi-language real-world software engineering benchmark. Resolves complex GitHub issues across Python, JavaScript, TypeScript, Rust, and more.
Builds fully functional, visually polished web apps and games from scratch. GPT‑5.3‑Codex autonomously iterated on complex games over millions of tokens, understanding intent without exhaustive specification.
Reads documentation, runs web research, analyzes structured data, and writes comprehensive reports. It can summarize thousands of data points in minutes with concise, actionable insights.
Matches GPT‑5.2 on GDPval — OpenAI's benchmark across 44 occupations. Creates presentations, spreadsheets, PRDs, financial analyses, training documents, and more at professional quality.
GPT‑5.3‑Codex achieves new state-of-the-art results on four benchmarks measuring coding ability, terminal proficiency, computer use, and professional knowledge work.
GPT‑5.3 Codex is best suited where coding, tool‑use, and structured reasoning must come together.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
GPT-5.3-Codex is documented as suited to debugging code and fixing failures in software workflows. It is also suited to deploying and monitoring software systems as part of the full software lifecycle. Another documented use case is writing product requirements documents, user research, and editing copy. It is also described as useful for slide decks and analyzing data in spreadsheets.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "openai/gpt-5.3-codex", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "openai/gpt-5.3-codex", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"openai/gpt-5.3-codex","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| OSWorld | 64.7% | Computer-use across real desktop applications | Source | July 12, 2026 |
| Terminal-Bench | 77.3% | Autonomous shell/terminal task completion | Source | July 12, 2026 |
| Intelligence | 32.5 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
GPT-5.3 Codex This page | Reasoning + agents | |||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
GPT-5.3 Codex has a 400,000 tokens context window and can return up to 128,000 tokens.
GPT-5.3 Codex takes image, text as input and returns text.
Use openai/gpt-5.3-codex as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
GPT-5.3 Codex became available on March 2, 2026.
GPT-5.3 Codex is priced at input $4.8139 / 1M tokens, output $38.5112 / 1M tokens, cached input $0.48139 / 1M tokens.
Yes, GPT-5.3 Codex can stream responses as they are generated.
Yes, GPT-5.3 Codex accepts image input alongside text.
GPT-5.3 Codex was built by OpenAI.
