OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026, the company's newest pair of flagship and everyday models. Astra takes over as the model for the hardest professional work, while Sol handles everyday coding and writing at a lower price. The launch arrived the same day in ChatGPT Work, Codex and the API, and it came with a price cut: Astra costs 20 percent less than the GPT-6 Astra it replaces.
Astra is the flagship. OpenAI positions it for the most demanding professional work, and it is priced at $8 per million input tokens and $40 per million output tokens through the API, down from the previous generation. Sol is the efficient option for everyday coding and writing, holding its price of $2 per million input tokens and $10 per million output tokens. Both models ship with a 1,050,000-token context window, with up to 922,000 tokens of input and 128,000 tokens of output in a single run.
Early independent benchmarks point upward. Artificial Analysis puts Astra at 62 on its Intelligence Index, up from 57 for GPT-6 Astra, and Sol at 57.3, up from 55. OpenAI also made its Ultrafast option available to Pro 200 subscribers for the first time through Sol, at half price for three months. The models are available the same day in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, as well as through the API as gpt-6.2-astra and gpt-6.2-sol.
The release also carried a notable safety disclosure. The system card rates Astra as the first externally released OpenAI model at the High level for AI self-improvement under the company's Preparedness Framework, and both models are rated Critical for cybersecurity. Those ratings do not block the release, but they are the kind of detail that shapes how enterprises and regulators talk about the models in the months ahead.
OpenAI has been shipping model updates at a fast cadence this autumn, and the 6.2 pair follows the GPT-6 generation that arrived earlier in the year. The pattern is familiar by now: a flagship model for the hardest work, a cheaper model for the everyday work, and a price cut that makes the flagship easier to run at scale.
A cheaper flagship model changes the economics of AI content production. When the best model costs 20 percent less to run, the teams already generating articles, product copy and outreach drafts at scale can generate more of it for the same budget, and the teams that were priced out can start. That means more AI-written content competing for the same search results, and more pressure on the content that is genuinely original, sourced and hard to replicate.
The bigger context window matters too. A model that can hold more than a million tokens can ingest an entire research corpus, a full competitor content set, or a long brief in one pass, and produce work that looks more thorough than what a smaller model could manage. For SEOs, the practical effect is that the bar for thoroughness keeps rising: a well-researched article is no longer a differentiator on its own, because the machines can now fake thoroughness convincingly.
There is also a citations angle. ChatGPT's answers draw on the models behind it, and a stronger default model changes which pages get surfaced and how confidently they are quoted. Content that is structured to be cited, with clear facts, named sources and direct answers, stands a better chance of being picked up no matter which model version is answering. The model name changes; the citation game does not.
Watch how quickly Astra reaches the free and Plus tiers of ChatGPT, and whether it becomes the default model behind ChatGPT's web search answers. Watch also for the first wave of benchmark results beyond Artificial Analysis, and for how rival labs respond on price. OpenAI has now cut its flagship price while raising its benchmark scores, and that combination puts pressure on Anthropic, Google and the rest to answer in kind.
Sources: 9to5Mac and WinBuzzer, November 17, 2026.
Related reading on this site: AI Overviews Showed Up in Only 20% of Local Searches, and Mostly Copied the Map Pack, ChatGPT Wallet Lets AI Agents Pay Without Seeing Your Card Number, November 2026 Core Update Still Rolling Out as Volatility Peaks on November 14, Domain Authority Barely Correlates With AI Citations, Search Atlas Finds, Social Networks Are a Minor AI Search Source — Except LinkedIn and YouTube, and Semrush Rebuilds Traffic and Market Data With a New Model, and Old Numbers Will Change.
Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.
Start free trial