OpenAI
Circuit Sparsity v1.0
Circuit Sparsity is a set of open tools developed by OpenAI to connect weight-sparse models and dense baselines through activation bridges. These tools aim to enhance the interpretability and efficiency of transformer models by enforcing sparsity during training, resulting in more compact and interpretable circuits.
Fence v0.9.1
Fence is a security tool designed to harden CI/CD pipelines by enforcing bounded egress filtering and runner lockdown within GitHub Actions. It allows developers to restrict network access, disable passwordless sudo, and manage container permissions to reduce the attack surface of automated workflows.
Hosted Shell Quickstart
The Hosted Shell Quickstart is a guide provided by OpenAI to help developers integrate and utilize the Hosted Shell feature within their applications. This feature allows for seamless execution of shell commands in a secure and controlled environment, enhancing the capabilities of AI agents by enabling them to interact with the system's shell directly.
Prism v1.0
Prism is a free, LaTeX-native workspace that integrates GPT‑5.2 directly into scientific writing and collaboration. It offers a unified platform for drafting, compiling, and collaborating on scientific documents, enhancing productivity and efficiency.
Daybreak
OpenAI's Daybreak is an advanced AI-driven solution designed to revolutionize cybersecurity by integrating proactive defense mechanisms directly into software development processes. It aims to enhance software resilience from the outset, enabling defenders to identify and address vulnerabilities more effectively.
GPT-4 Turbo v4-turbo
The latest and most advanced GPT-4 model with improved performance, longer context window of 128K tokens, and lower pricing. Optimized for chat, code generation, and complex reasoning tasks.
GPT-5.4 mini and nano
OpenAI has introduced GPT-5.4 mini and nano, advanced models optimized for high-volume workloads, offering enhanced performance in coding, reasoning, and multimodal understanding. These models are designed to deliver faster and more efficient responses, making them ideal for applications requiring rapid processing and cost-effectiveness.
GPT-5.6 Sol, Terra, and Luna v5.6
OpenAI introduces the GPT-5.6 series of large language models (LLMs). The GPT-5.6 series introduces a new tiered naming system designed to give developers and users clearer choices, balancing intelligence, speed, and cost. GPT-5.6 Sol: The flagship model. It is the most capable model yet, featuring advanced agentic capabilities, deep reasoning (with a new max reasoning effort and ultra mode), and top-tier performance in coding, biology, and cybersecurity. GPT-5.6 Terra: The balanced model. It delivers everyday performance competitive with GPT-5.5, but at half the cost. GPT-5.6 Luna: The lightweight model. It is the fastest and most affordable option in the lineup, bringing strong capabilities at the lowest price point.
GPT-5.3-Codex v5.3
GPT-5.3-Codex is OpenAI's most advanced agentic coding model, combining the coding performance of GPT-5.2-Codex with the reasoning and professional knowledge capabilities of GPT-5.2. It enables long-running tasks involving research, tool use, and complex execution, allowing interactive steering without losing context.
GPT-5.5 v5.5
GPT-5.5 is OpenAI's latest AI model, designed to enhance productivity by efficiently handling complex tasks such as code writing and debugging, online research, data analysis, and document creation. It excels in agentic coding, computer use, knowledge work, and early scientific research, offering a significant boost in intelligence without compromising speed.
Codex Security Review vResearch Preview
Codex Security Review is an AI-powered security analysis tool designed to perform in-depth evaluations of GitHub pull requests. It analyzes code diffs, repository context, and custom threat models to identify and report security vulnerabilities with high precision.
Symphony
Symphony is an open-source project developed by OpenAI that transforms project work into isolated, autonomous implementation runs. This approach enables teams to manage tasks more effectively by reducing the need to supervise coding agents.
Codex Security v0.1.1
Codex Security is an application security agent designed to help development teams identify, validate, and remediate vulnerabilities within their codebases. It provides both a CLI and a TypeScript SDK to scan repositories, review changes, and propose actionable fixes for security issues.
OpenAI Frontier
OpenAI Frontier is a platform designed to help enterprises build, deploy, and manage AI agents capable of performing real-world tasks. It provides agents with essential skills such as shared context, onboarding, hands-on learning with feedback, and clear permissions and boundaries, enabling teams to transition from isolated use cases to AI coworkers operating across the business.
Codex Agent-Native Platform v2026
OpenAI re-launched Codex as a standalone agent-native platform rather than just a model. It targets deterministic execution for complex, multi-file repository refactors with high follow-through, enabling developers to automate intricate engineering workflows with predictable, reliable outcomes.
GPT-5.3-Codex-Spark v5.3-Codex-Spark
GPT-5.3-Codex-Spark is an ultra-fast model designed for real-time coding tasks, optimized to deliver over 1000 tokens per second. It is a smaller version of GPT-5.3-Codex, tailored for rapid, interactive coding experiences.
Developer Mode v26.609
OpenAI's Codex app introduces an in-app browser with Developer Mode, enabling developers to preview and comment on web applications directly within the Codex environment. This feature facilitates efficient debugging and development by allowing real-time interaction with web pages without the need for external browsers.
OpenAI Advanced Account Security
A critical infrastructure update for the "agentic" era. It introduces mandatory passkeys and recovery keys for high-privilege accounts to prevent unauthorized access as GPT-5.5 begins performing more autonomous, real-world tasks.