Unveiling the Realtime API
This article explains how speech-to-speech models replace traditional multi-stage voice pipelines (ASR, LLM, TTS) with a single model that processes raw audio directly, dramatically reducing latency a…
Coding with AI
AI coding assistants and agents, testing and debugging with AI, and practical setups for developers.
MLE-bench: A Benchmark for Assessing AI Agents in Machine Learning Engineering Tasks
MLE-bench is a benchmark created to evaluate the effectiveness of AI agents in performing machine learning engineering tasks, measuring their capability across real-world ML engineering scenarios.
Presenting the SWE-Lancer benchmark
The Remote Labor Index benchmark evaluates whether frontier AI models like GPT-5 and Claude Sonnet 4.5 could earn $1 million completing real freelance coding jobs from Upwork-style platforms. Results …
Supplement to the o3 and o4-mini system card: Codex
Codex is a cloud-based coding agent from OpenAI that handles complete development tasks—bug fixes, features, tests—while developers work in parallel. Powered by codex-1, an o3 model variant optimized …
Shipping code more quickly using o3, o4-mini, and GPT-4.1
CodeRabbit is an AI-powered code review assistant that uses OpenAI models like o3, o4-mini, and GPT-4.1 to scan pull requests instantly, catch bugs, and speed up merges. Acting as a first-pass reviewe…
GPT-5 Technical Documentation
GPT-5 uses a unified routing system that automatically directs each request to the most suitable model variant. gpt-5-main handles everyday conversations quickly, gpt-5-thinking tackles complex reason…
Announcing new improvements to Codex
OpenAI's Codex update delivers major gains in speed and reliability, enhanced real-time collaboration, and stronger autonomous task handling. Available across terminal, IDE, web, and mobile, Codex now…
Developing an approach for estimating age
OpenAI is introducing age prediction technology and expanded parental controls in ChatGPT to make the AI assistant safer for teenagers. The system estimates user age from usage patterns, applies stric…
Cyber Operations: Support for Phishing Campaigns and Script Development
OpenAI has terminated multiple accounts whose behavior matched known threat groups and aligned with Chinese intelligence priorities. These accounts exploited OpenAI's AI tools to support phishing camp…
Cyber Campaign: Malware toolsets used by Russian-speaking threat actors
OpenAI banned accounts linked to Russian-speaking cybercriminal networks that used its AI tools to develop malware loaders, obfuscation layers, credential-stealing scripts, and command-and-control inf…
Announcing AgentKit, fresh Evals, and RFT for agents
OpenAI releases three key resources to help developers move from prototype to production faster: AgentKit, a unified toolkit for building, testing, and shipping agentic applications; expanded Evals fe…
Achieving more with GPT-5.1-Codex-Max
GPT-5.1-Codex-Max is OpenAI's newest agentic coding model designed specifically for Codex. Unlike chat assistants, it works autonomously on long-running, project-scale tasks such as refactoring and fr…
GPT-5 and what lies ahead for mathematical breakthroughs
UCLA math professor Ernest Ryu collaborated with OpenAI's GPT-5 to address a significant open question in optimization theory, the mathematics underlying modern machine learning. The partnership highl…
App submissions for ChatGPT are now open to developers
OpenAI has officially opened app submissions for ChatGPT, ending the experimental phase. Approved apps will be listed in a new in-product directory, giving developers direct access to ChatGPT's massiv…
Strengthening ChatGPT Atlas against prompt injection attacks on an ongoing basis
OpenAI's ChatGPT Atlas browser can click, type, and navigate on users' behalf, making it a target for prompt injection attacks where hidden web page instructions trick the AI into harmful actions. Ope…
Datadog leverages Codex for system-wide code review
Datadog has adopted OpenAI's Codex to enable system-wide code review, leveraging AI to automatically analyze code changes across its platform. This integration helps improve code quality, catch issues…
Unpacking the Codex agent loop
This article explains how the Codex CLI agent loop works under the hood. It covers the iterative cycle of receiving context, deciding on actions, executing tools, and observing results. It highlights …
Meet the Codex app
OpenAI has launched the Codex app for macOS, a native hub for AI-assisted software development. It lets developers run multiple coding agents simultaneously, manage parallel workflows, and handle long…
Say Hello to GPT-5.3-Codex
GPT-5.3-Codex is a Codex-native AI coding agent built from the ground up for long, real-world engineering tasks. Unlike chatbots adapted for code, it combines frontier-level coding performance with ge…
Inside the Codex harness: building the App Server
The Codex App Server lets developers embed an AI coding agent directly into their applications via a bidirectional JSON-RPC API. Unlike one-shot prompt services, it supports ongoing conversations wher…