Update: A day later, GPT-6 Astra will launch for Business and Pro customers with the $100/month or $200/month plan. This follows further launches of new models before Plus customers on the $20/month plan gain access. It’s been just over a year since OpenAI released the first GPT-5 models for ChatGPT. Now it’s time for GPT-6 Astra to update ChatGPT and Codex capabilities. Details below. GPT-6 Astra Details OpenAI has officially detailed its new model, GPT-6 Astra, which it calls “the world’s most intelligent and aligned model.” Here’s more about today’s official announcement: GPT-6 Astra brings together years of research and big bets on pre-training, reinforcement learning, and alignment. Astra is at the forefront of computer use, navigation, software engineering, cybersecurity, science and professional work. Astra saturates FrontierMath Tier 4 with a score of 98%, and has already helped solve long-standing open problemsin mathematics. Astra also saturates ARC-AGI-3 with a score of 99.9% and ExploitBench with a score of 100%. It also sets a new frontier in computer and browser use, handling the most demanding professional work with unmatched speed, precision, and judgment. OpenAI goes further in how GPT-6 improves computer usage, allowing ChatGPT to use your Mac or other desktop, even in the background while you work: GPT-6 Astra combines advances in computer usage with training specific to professional environments, to help tackle complex work tasks. It combines the intelligence needed for complex problems with the ability to carry out multi-step workflows and produce polished documents, spreadsheets and presentations. Additionally, GPT-6 Astra is “the best software engineering model to date,” says OpenAI. With Astra, ChatGPT is almost 2x faster on the computer than before. In addition to the incredible new model, we have optimized the harness to make computer use much faster. These optimizations also work on existing models, so tasks are speeded up by about 60% on GPT-5.6 Sol. Try it! pic.twitter.com/2EXH5f0uZL — Ari Weinstein (@AriX) September 4, 2026 In Codex-specific terms, here’s something new about GPT-6: with Astra, we’re introducing a new way for Codex to preserve and recover context when the context window fills up. Historically, models have used compaction to summarize work over long sessions, such as when debugging complex problems or tackling large refactorings. Each compaction can omit details about why a solution failed or how a component behaves. In Codex, Astra can save notes in contextual windows, preserving accumulated details without repeatedly compressing them into a single summary. Previous context windows are still searchable, so Astra can find requirements or test results from previous messages and tool results, even if that information was not captured in your notes. You can enable this experimental feature in your Codex and it will become the default for Astra in the coming weeks. We launched something… it is best in class in computer use, scientific research, coding, mathematics, professional work like making slides and much more. He’s superhuman at absorbing enormous amounts of context. so patient so relentless. very good at changing context. and it will… https://t.co/l3yM1Sy68n — Kundan Kumar (@kundan2510) September 3, 2026 OpenAI calls the GPT-6 Astra a “significant leap in cyber capabilities,” adding that it “meets the critical thresholdin cybersecurity under our Readiness Framework.” For that reason, the model is being slowly rolled out to users. “GPT-6 Astra is rolling out today to a limited set of organizations and in the coming days will be available to all ChatGPT Plus, Pro, Business and Enterprise users, as well as through the OpenAI and AWS API,” OpenAI says. “Use of Astra is included within existing subscription allowances – users and businesses will also be able to purchase credits for additional usage. Users on Pro, Business and Enterprise plans will also have access to GPT-6 Astra Pro. Enterprise administrators can enable Astra for their workspace; access is disabled by default at launch.” The Road to GPT-6 Astra GPT-5 was released on August 7, 2025, replacing GPT-4o and various reasoning models as ChatGPT’s default. Its defining feature was automatic routing: ChatGPT could respond quickly or invoke deeper reasoning depending on the request. OpenAI also claimed fewer factual errors than GPT-4o. Apple adopted the model with iOS 26, iPadOS 26, and macOS Tahoe 26. ChatGPT requests passed through Apple Intelligence, including Siri transfers, writing tools, and visual intelligence, and changed from GPT-4o to GPT-5. GPT-5.1 promised warmer, more natural responses in November. December’s GPT-5.2 “code red” accelerated professional work in response to Gemini. GPT-5.3 Instant focused on ChatGPT’s “embarrassing” tone, followed two days later by GPT-5.4 Thinking with a million-token context window and built-in computer usage. Later, OpenAI added GPT-5.4 mini and nano, the GPT-5.4-Cyber restricted, GPT-5.5 for more autonomous work, GPT-5.5 Instant and the GPT-5.6 Sol, Terra and Luna models from July. From the Codex app for Mac to a new ChatGPT. The software related to those models was also transformed. OpenAI launched a dedicated Codex Mac app in February as a hub for running multiple coding agents, reviewing changes, and scheduling work. In July, Codex effectively became the new ChatGPT desktop app, combining Chat, Work, and Codex, while the previous native client was renamed ChatGPT Classic. For me, Codex has changed the way I work with ChatGPT the most. I can offload a lot more work that doesn’t require critical thinking and even create tools on my Mac that I wouldn’t otherwise be able to create. GPT-6 Availability OpenAI says that to start, its new GPT-6 models will only be available to members of its cybersecurity-focused Daybreak program. This reflects a similar launch cadence to the previous model, GPT-5.6. Here’s what OpenAI says specifically about availability: GPT-6 Astra is rolling out today to a limited set of organizations and in the coming days will be available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as via the OpenAI API and AWS. You can learn a lot more about OpenAI’s latest frontier model here. FTC: We use automatic affiliate links that generate income. Further.
