Tag
All content about OpenAI, organized for fast scanning.
OpenAI has designated its upcoming Astra model as a "critical" cybersecurity tool, emphasizing the importance of implementing controls before its release. Additionally, the company has announced significant price reductions for its GPT-5.6 models and introduced new features in its ChatGPT desktop app, including voice control and integration with Codex, enhancing user experience and functionality. These developments reflect OpenAI's ongoing efforts to improve accessibility and performance across its AI offerings.
OpenAI has classified its upcoming Astra as its first “critical” cybersecurity model under its Preparedness Framework. The company says it’s adding controls before continuing development, while aiming for broad availability focused on defenders.
OpenAI is rolling out pricing and speed updates for the GPT-5.6 lineup, including 80% lower Luna prices and 20% lower Terra rates. GPT-5.6 Sol gets faster /fast mode, while auto-approve “review for me” is now ~10x cheaper using Luna.
OpenAI has just rolled out ChatGPT Voice in its desktop app, bringing hands-free control and multi-agent direction via GPT-Live. It’s launching globally on macOS and Windows for paid plans, with iOS Codex support and Android “coming soon.”
With the launch of the new ChatGPT desktop app, OpenAI is folding Codex into a dedicated workspace alongside a revamped browser, faster Computer Use, and a new ChatGPT Work Chrome extension. Sites also enters beta for Plus and Pro, while users ask about Windows, Linux, and existing workflows.
Anthropic says Claude Fable 5’s new classifiers will send a small fraction of routine coding and debugging back to Opus 4.8. The fallback will be transparent, billed at Opus rates, and aims to reduce false positives as safeguards are refined.
Zai.org has just rolled out GLM-5.2, billed as “Frontier Intelligence, Open Weights.” The announcement points to major gains in coding and agentic tasks, with another promised strength only partially visible. A repost by Jiayuan Zhang quickly drew over 1,100 shares.
OpenAI Codex lead Thibault Sottiaux reminded developers that the Codex app, CLI, and SDK can work with “any open source model,” not just OpenAI’s.
OpenAI is rolling out more Codex features across the EEA, UK, and Switzerland this week, including Computer use, the Chrome extension, memory, and Chronicle. Users welcomed the news but raised questions about platform support, limits, and capacity.
OpenAI has just rolled out a new “Developer mode” for Codex browser use, enabling Chrome DevTools Protocol access to profile JS performance and inspect console, network, and page state. Settings flag “Elevated risk,” warning full CDP access can expose sensitive browser internals.
OpenAI has just rolled out Lockdown Mode for ChatGPT, limiting web-connected features to reduce prompt-injection risk. It’s also standardizing “Elevated Risk” labels across ChatGPT, Atlas, and Codex to flag network-exposed capabilities.
Antigravity has just rolled out /teamwork-preview for all paid plans, bringing parallel implementation and verification agents for complex tasks. Varun Mohan says it’s a research preview that can burn through tokens—and claims it’s already built a working OS.
OpenAI has just rolled out Codex “Sites,” bringing a new way to turn work and ideas into interactive websites or apps shared by URL. It’s launching first for Business and Enterprise users, as developers debate where it fits and how it handles security, auth, and scale.
OpenAI has just rolled out general availability for its frontier models and Codex on AWS. The move lets enterprises tap existing Bedrock security, compliance, and governance workflows—though some users say the models weren’t yet showing up in catalogs.
OpenAI has just rolled out Daybreak, pitching it as “frontier AI for cyber defenders” that combines its top models, Codex, and security partners. The company says it can help find and fix vulnerabilities earlier, but details and access remain unclear.
OpenAI’s latest post explains how it’s deploying Codex with tighter sandboxing, approval gates, and policy controls to reduce risky actions. It also highlights deep logging via OpenTelemetry so teams can audit prompts, tools, and network activity.
OpenAI has just rolled out a new Codex Chrome plugin designed to boost browser-based dev work, from testing web apps to gathering context across tabs. Early reactions are mixed, with praise for the workflow and questions about auth, profiles, and reliability.
OpenAI has just rolled out GPT-5.5-Cyber, bringing a new “frontier cybersecurity model” to critical cyber defenders. Early reactions focus on trusted access, benchmarks, and whether controls can prevent misuse by attackers.
OpenAI is expanding its partnership with AWS with three new Bedrock integrations in limited preview: OpenAI models (including GPT-5.5), Codex, and Bedrock Managed Agents powered by OpenAI. The goal is smoother enterprise deployment with familiar AWS security, identity, and billing.
OpenAI has just rolled out Symphony, an open-source spec and reference implementation for orchestrating coding agents around project boards like Linear. It assigns an agent per issue, keeps work running with CI awareness, and claims big PR gains—though the metric deserves scrutiny.
OpenAI has just rolled out its GPT-5.5 System Card, detailing how the new model targets complex, multi-tool work like coding, research, and analysis. The report also outlines pre-deployment testing, red-teaming, and “strongest safeguards to date,” plus notes on GPT-5.5 Pro.
OpenAI has just rolled out GPT-5.5, bringing bigger gains in agentic coding, knowledge work, and computer use while matching GPT-5.4 latency. It’s arriving first in ChatGPT and Codex, with API access “very soon” and new pricing tiers.
Ramp Labs says coding agents blow past budgets even with live meters and explicit approvals. In SWE-bench tests, agents almost always chose to keep spending, and separate “controller” models were easily swayed by bad recommendations.
OpenAI has just rolled out Chronicle for Codex, enhancing its memories with recent on-screen context to cut down on repeated prompts. It’s launching to Pro users on macOS (excluding the EU, UK, and Switzerland), as developers weigh usefulness against privacy and security risks.
OpenAI has just rolled out a major Agents SDK update focused on the agent harness and secured sandbox execution. New memory and filesystem tools, plus a portable workspace manifest, aim to make long-running, tool-using agents more reliable in production.
With the launch of GPT-5.4-Cyber, OpenAI is rolling out a more cyber-permissive model for vetted defenders. The company is also scaling Trusted Access for Cyber to thousands of verified individuals and hundreds of teams protecting critical software.