OpenAI has launched two new artificial intelligence models, GPT-6.2 Astra and GPT-6.2 Sol, designed for demanding professional work and everyday tasks respectively, with the flagship Astra model costing 20 percent less than its predecessor. The models were released on Monday, 17 November 2026, in ChatGPT Work and Codex for paid users, and to developers through the OpenAI API.
Key Takeaways
- GPT-6.2 Astra is OpenAI’s new flagship for complex professional work; GPT-6.2 Sol is a faster, cheaper model for everyday multi-step tasks.
- Astra costs $8 per million input tokens and $40 per million output tokens — 20 percent below GPT-6 Astra — while Sol costs $2 and $10, unchanged from GPT-6.1 Sol.
- Both models share a 1,050,000-token context window with up to 128,000 tokens of output.
- The models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and via the API as gpt-6.2-astra and gpt-6.2-sol. They are not in regular Chat and not available to free accounts.
- OpenAI says Astra improves on its predecessor by 6.2 percentage points on its GDPval benchmark while using 22 percent fewer output tokens.
Two models, two jobs
OpenAI positions the two models for different kinds of work. GPT-6.2 Astra is built for demanding professional tasks: preparing documents, presentations and spreadsheets, writing code, carrying out scientific research, financial analysis and legal research. GPT-6.2 Sol is intended for everyday work that involves multiple steps — routine coding, document writing and tasks that require the model to use tools and make decisions along the way. The naming is slightly unusual: GPT-6.2 Astra is an upgrade to GPT-6 Astra, while GPT-6.2 Sol is built on GPT-6.1 Sol, so it upgrades that earlier model rather than a GPT-6 Sol. Both models accept text and images and return text, and both work in ChatGPT Work — the business and professional environment — rather than in the regular ChatGPT chat used by consumers. Developers can call them through the Responses API and Chat Completions as gpt-6.2-astra and gpt-6.2-sol, with tool calling available through the Responses API.
Lower prices, longer context
The headline change for businesses is price. Astra now costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for GPT-6 Astra — a 20 percent cut — and cached input costs are 60 percent lower. Sol’s pricing is unchanged from GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, making Sol exactly a quarter of Astra’s price. Both models share the same context window of 1,050,000 tokens, with a maximum output of 128,000 tokens, so they can work with very long documents and codebases in a single session. Their knowledge cutoffs differ slightly: Astra’s training data runs to 18 May 2026, while Sol’s runs to 1 June 2026. OpenAI is also extending its fastest tier, Ultrafast: Pro 500 subscribers can use Ultrafast with either model, and for the first time Pro 200 subscribers get Ultrafast too, starting with GPT-6.2 Sol. Ultrafast is priced at three times the standard rate through 28 February 2027, rising to six times from 1 March 2027.
What the benchmarks say
OpenAI’s own launch material reports solid gains. The company says Astra improves on GPT-6 Astra by 6.2 percentage points on GDPval v2 — a benchmark of real professional knowledge work — while using 22 percent fewer output tokens, which is what drives the lower cost per task. Sol improves by 5.3 points over GPT-6.1 Sol on the same measure. Astra also jumps 22 points over GPT-6 Astra on ExploitBench Internal Port, a cybersecurity benchmark, and OpenAI describes Astra as its most capable model ever for cyber work. Independent researchers have begun testing the models too: Epoch AI, the research group that tracks AI progress, placed Astra at 174 on its Capabilities Index, level with Anthropic’s Claude Fable 5.2 and above the trend line that reasoning models had followed before Astra existed. On a robotics benchmark run through Codex, Artificial Analysis made Astra the new leader. These early figures are vendor-reported or first-day independent tests, and wider independent evaluation will follow in the coming weeks — but the direction is consistent: Astra is a step up in capability, and Sol closes much of the gap at a quarter of the price.
Stronger safeguards for cyber capabilities
Because Astra is significantly more capable in cybersecurity, OpenAI says it is launching both models with tighter safeguards for higher-risk cybersecurity assistance. The system card published with the launch assesses both models at OpenAI’s High threshold for biological and chemical capabilities and at its Critical threshold for cybersecurity, while Astra also reaches the High threshold for AI self-improvement and Sol does not. The safeguards reflect a broader industry pattern: as models become better at finding software vulnerabilities and automating complex technical work, their makers restrict the most sensitive uses. OpenAI has also changed how the API handles reasoning in this release: starting with GPT-6.2, the API drops earlier reasoning items when it sees that the conversation history before them was edited, or that the safety identifier changed between requests — a change developers upgrading from earlier models will need to account for.
What it means for users and developers
For businesses already using ChatGPT Work or Codex, the practical effect is straightforward: the most capable model is now cheaper per task, and the everyday model is nearly as capable for routine work at a quarter of the price. Microsoft, which offers the models through its own platform, says Astra delivers higher average accuracy than GPT-6 Astra with substantially fewer output tokens, and highlights improvements in reasoning, writing, coding, computer use and creating documents, slides and spreadsheets. For Sol, Microsoft points to more concise technical writing, more focused coding and debugging, better decisions about when and how to use tools, and front-end design. The launch comes as competition among frontier AI labs intensifies, with each new release judged on capability per dollar as much as on raw benchmark scores. With Astra cheaper and Sol close behind it on the early independent benchmarks, OpenAI is betting that efficiency — doing the same work with fewer tokens — matters as much to customers as raw power. The models are rolling out now, though availability depends on plan, workspace permissions and rollout progress, so some users may not see them immediately.
Frequently Asked Questions
What are GPT-6.2 Astra and Sol?
They are two new AI models from OpenAI, launched on 17 November 2026. Astra is the flagship for demanding professional work such as documents, spreadsheets, coding and research, while Sol is a faster, cheaper model for everyday multi-step tasks.
How much do GPT-6.2 Astra and Sol cost?
Astra costs $8 per million input tokens and $40 per million output tokens, 20 percent less than GPT-6 Astra. Sol costs $2 per million input tokens and $10 per million output tokens, unchanged from GPT-6.1 Sol.
Who can use GPT-6.2 Astra and Sol?
They are available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and to developers via the OpenAI API. They are not available in regular Chat or to free accounts.
Sources:
- OpenAI, “GPT-6.2 Astra and Sol are now available”, 17 November 2026
- The Economic Times (ETCIO), “OpenAI Unveils Cost-Effective GPT-6.2 Astra and Sol for Enhanced Professional Efficiency”, 18 November 2026
- Digital Today, “OpenAI releases GPT-6.2 Astra and Sol for ChatGPT Work and Codex”, 18 November 2026

Pingback: GTA 6 Release Date: Platforms, Pre-Load and Launch Facts
Pingback: GTA 6 Price and Editions: Standard vs Ultimate Explained
Pingback: Google Launches Gemini 4 Neon AI Model at $3/$15
Pingback: GAO: ChatGPT, Claude, Gemini, Grok Produced Terror Plans