OpenAI GPT-6.2 Astra & Sol: Price, Features & Access

Published

The quick answer
OpenAI launched two new models on November 17, 2026: GPT-6.2 Astra and GPT-6.2 Sol.
- Astra is the frontier model. It is built for the hardest professional work.
- Sol is the efficient everyday model. It is built for production agents that run all day.
- Both are available from day one in ChatGPT Work and Codex, and through the OpenAI API.
- Astra costs $8 per million input tokens and $40 per million output tokens through the API.
- Sol costs $2 per million input tokens and $10 per million output tokens through the API.
- Both offer a 1,050,000-token context window and up to 128,000 output tokens.
- Astra is 20% cheaper than GPT-6 Astra on API input and output pricing.
- Sol pricing is unchanged from GPT-6.1 Sol.
- Ultrafast inference is available for both models at launch.
- Astra scores 62 on the Artificial Analysis Intelligence Index, up from 57 for GPT-6 Astra.
If you want the strongest model for demanding work, start with Astra. If you want a lower-cost model for agents that run all day, start with Sol.
What OpenAI launched on November 17, 2026
OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol on the same day. The launch covers two models, not one. Astra sits at the top of the range. Sol sits below Astra on price and is aimed at everyday production use. Both models launched together in ChatGPT Work, Codex, and the API. That shared launch date matters for teams planning a switch. You do not need to wait for a second rollout to try either model. The two models are designed to work as a pair. Astra handles the hardest tasks. Sol handles high-volume, always-on work. Together, they cover professional work and everyday agent work.
Astra vs Sol at a glance
| Feature | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
| Role | Frontier model for the hardest professional work | Efficient everyday model for production agents |
| API input price | $8 per million tokens | $2 per million tokens |
| API cached input price | $0.40 per million tokens | $0.10 per million tokens |
| API output price | $40 per million tokens | $10 per million tokens |
| Context window | 1,050,000 tokens | 1,050,000 tokens |
| Maximum output | 128,000 tokens | 128,000 tokens |
| Knowledge cutoff | 18 May 2026 | 1 June 2026 |
| Ultrafast inference | Available at launch | Available at launch |
| Best for | Demanding reasoning, writing, coding, and documents | Agents that run all day |
Key points: Astra costs more per token. Sol costs much less per token. Both share the same context window size and maximum output size. The main difference is the intended workload, not the context size. Astra is positioned for work where quality matters most. Sol is positioned for work where volume and cost matter most.
GPT-6.2 API pricing in detail
API pricing is charged per million tokens. Input tokens are the tokens you send to the model. Output tokens are the tokens the model generates. Cached input tokens are a lower-cost category for repeated input. Astra and Sol use the same pricing structure, with different rates for each model. Astra is the higher-priced model. Sol is the lower-priced model.
| Model | Input per million tokens | Cached input per million tokens | Output per million tokens |
|---|---|---|---|
| GPT-6.2 Astra | $8 | $0.40 | $40 |
| GPT-6.2 Sol | $2 | $0.10 | $10 |
| GPT-6.1 Sol (previous) | $2 | $0.10 | $10 |

Astra is 20% cheaper than GPT-6 Astra. That 20% reduction applies to the API pricing comparison with GPT-6 Astra. Cached input for Astra is 60% cheaper than before, now $0.40 per million tokens β a large saving for repeated prompts and long documents. If you reuse the same context often, cached input matters. Sol pricing is unchanged from GPT-6.1 Sol.
What the pricing means in practice: Astra costs four times as much as Sol for input, output, and cached input. That gap is the price of frontier capability. Sol is the better choice when the task does not need the strongest model. Astra is the better choice when the task is hard and the result matters. Many teams will use both: hard tasks to Astra, routine tasks to Sol.
Long-context pricing note: Above 272,000 input tokens, input and cache rates double and output rises to 1.5x for the whole request. This applies to both Astra and Sol. Check your input size before you send a very long request. A request just above the threshold costs more per token. This rule matters most for long documents and large codebases. Plan your prompts and context around the 272,000-token line. Split very large jobs when you can. Cache repeated input when you can.
Context window and output limits
Both models offer a very large context window of 1,050,000 tokens β just over 1 million tokens. It allows very long documents, large codebases, and long conversations in a single request. Maximum output is 128,000 tokens for both models, supporting long reports, long code files, and long documents. You do not need to choose between the models for context size.
Knowledge cutoff dates: Astra 18 May 2026; Sol 1 June 2026. Sol has the later cutoff by about two weeks. Neither cutoff covers events after its date. Check recent facts separately if your work depends on them.
What Astra does better
OpenAI positions Astra as the stronger model for demanding work: reasoning, writing, coding, computer use, creating documents, slides and spreadsheets, and building websites, games and 3D designs. That list covers a wide range of professional work. If your work touches several of these areas, Astra is the natural first choice. If your work touches only one simple area, Sol may be enough.
Where you can use GPT-6.2 from day one
Both models are available from day one in ChatGPT Work and Codex, and through the OpenAI API. Plus and Pro users get access from day one. Enterprise and Edu users get access from day one, once a workspace admin enables the models. Check with your workspace admin if you do not see them. The API model names are gpt-6.2-astra and gpt-6.2-sol β use those exact names in API calls. The API route suits apps, agents, and automated workflows, while ChatGPT Work and Codex suit direct, hands-on work.

Access checklist: Plus β available from day one in ChatGPT Work and Codex. Pro β available from day one in ChatGPT Work and Codex. Enterprise β available from day one, once a workspace admin enables it. Edu β available from day one, once a workspace admin enables it. Developers β available from day one through the OpenAI API.
Ultrafast inference at launch
Ultrafast inference is available for both models at launch. Pro 500 subscribers get it with either model. Pro 200 subscribers get it first with Sol. If speed matters, check which Pro tier you are on. If you are on Pro 200, Sol is your first ultrafast option. If you are on Pro 500, you can use ultrafast inference with either model.
Benchmark: Artificial Analysis Intelligence Index
Artificial Analysis scores GPT-6.2 Astra at 62 on its Intelligence Index. That is an independent benchmark score, not an OpenAI score. GPT-6 Astra scored 57 on the same index β 5 points lower. A higher score means stronger performance on that benchmark, but it does not mean Astra is better at every task. Use it alongside the pricing and feature details above.
OpenAI's self-improvement rating for Astra
OpenAI rates Astra as its first externally released model rated High for AI self-improvement capability. This is OpenAI's own rating, not an established fact about how the model behaves. Treat it as OpenAI's assessment, not as independent proof. It does not change the pricing or access details above, or which model to pick for everyday work.
Token allowance change coming in December 2026
From December 2026, Plus, Pro and Business plans will share one token-based allowance that covers ChatGPT and Codex together. It will not be a separate allowance for each product. Plan your ChatGPT and Codex use together from December. Heavy use in one product will affect the shared allowance.
Also available in Microsoft Foundry
GPT-6.2 is also generally available the same day in Microsoft Foundry, giving teams on Microsoft's platform a direct route. Related context: OpenAI also announced its frontier models and Codex are coming to Google Cloud, with general availability in the coming weeks.
Which model should you choose?
Choose Astra if: your work is hard and the result matters; you need strong reasoning, writing or coding help; you need documents, slides, or spreadsheets; you are building websites, games, or 3D designs; you need computer use for complex tasks; you are willing to pay more per token for stronger results.
Choose Sol if: you run production agents all day; you need an efficient everyday model; your tasks are routine or high-volume; cost per token matters more than maximum capability; you want the lowest API price in this launch; you are on Pro 200 and want ultrafast inference first; you want the same large context window at a lower price.
Use both if: you run a mix of hard and routine work; you want Astra for the hardest tasks and Sol for everything else; you want to control costs without losing quality; you build agents that can route tasks by difficulty.
What this means for developers and professionals
Developers get both models through the API on day one. Use gpt-6.2-astra for the frontier model and gpt-6.2-sol for the efficient model. Both share the same context and output limits, which makes switching simpler. Test the same prompt on both models, then compare the result and the cost. Route by task type in your code: hard tasks to Astra, routine tasks to Sol. Cached input is much cheaper for Astra than before, which helps apps that reuse the same instructions or documents. Watch the 272,000-token threshold on long requests, where per-token cost rises for the whole request.
For professionals, Astra is aimed at the hardest work: reasoning, writing, coding, documents, slides, spreadsheets, websites, games, and 3D designs. If your work is complex, test Astra first. If it is routine, Sol may handle it at a lower cost. For everyday users, Sol offers the same large context window at a lower price. Try Sol first for routine tasks, and Astra when a task is hard or important.
Things to check before you switch
Check whether your workspace admin has enabled the models, if you are on Enterprise or Edu. Check your Pro tier if you want ultrafast inference on Astra. Check your input size against the 272,000-token pricing threshold. Check your cached input use, because Astra cached input is now much cheaper. Check the December 2026 token allowance change if you are on Plus, Pro, or Business. Check recent facts separately, because Astra's cutoff is 18 May 2026 and Sol's is 1 June 2026. Test both models on your real tasks before you commit. Compare the result, the speed, and the cost. Do not assume the higher-priced model is best for every task. Do not assume the lower-priced model is too weak for your work.
Bottom line
GPT-6.2 gives you a clear choice, not a single new model. Astra is the frontier model for the hardest professional work, at $8 input and $40 output per million tokens. Sol is the efficient everyday model for agents that run all day, at $2 input and $10 output per million tokens. Both offer a 1,050,000-token context window and up to 128,000 output tokens. Both are available from day one in ChatGPT Work, Codex, and the API. Astra is 20% cheaper than GPT-6 Astra, with cached input 60% cheaper. Sol pricing is unchanged from GPT-6.1 Sol. Astra scores 62 on the Artificial Analysis Intelligence Index, up from 57. Ultrafast inference is available for both at launch. From December 2026, Plus, Pro and Business plans will share one token-based allowance across ChatGPT and Codex. Start with the model that matches your work. Use Astra for the hardest tasks. Use Sol for high-volume everyday work. Use both when your workload needs both.
FAQs
When did OpenAI launch GPT-6.2 Astra and Sol?
OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026.
What is the difference between GPT-6.2 Astra and GPT-6.2 Sol?
Astra is the frontier model for the hardest professional work. Sol is the efficient everyday model for production agents that run all day.
How much does GPT-6.2 cost through the API?
Astra costs $8 per million input tokens, $0.40 per million cached input tokens, and $40 per million output tokens. Sol costs $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens.
Where can I use GPT-6.2 Astra and Sol?
Both models are available from day one in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and through the OpenAI API as gpt-6.2-astra and gpt-6.2-sol. Enterprise and Edu users need a workspace admin to enable them.
What is the context window for GPT-6.2?
Both Astra and Sol have a 1,050,000-token context window and up to 128,000 output tokens.
What is GPT-6.2 Astra's benchmark score?
Artificial Analysis scores GPT-6.2 Astra at 62 on its Intelligence Index, up from 57 for GPT-6 Astra.
Was this article helpful?

Written By
TechorCoffeeContributing Writer
TechorCoffee is a technology reporter and software creator writing reviews and tutorials for TechOrCoffee.
Comments (0)
Leave a Comment
Read Next

Tue Apr 21 2026
Best AI Tools for Students & Creators in 2026:Complete Guide
AI tools are transforming how students learn and creators produce content. From writing assistants to design and research tools, these platforms save time, enhance creativity, and improve efficiency. Choosing the right AI stack can significantly boost productivity and output quality.

Tue Jun 16 2026
Apple Intelligence 2.0: Siri AI & iOS 27 at WWDC 2026
Apple Intelligence 2.0 is here. I cover Siri AI, iOS 27 release dates, supported devices, and how Apple's new AI really compares to ChatGPT and Gemini.

Tue Oct 13 2026
Gemini 4 Argon: Everything You Need to Know
Gemini 4 Argon is Googleβs new flagship AI model. See its benchmarks vs GPT-6 Astra and Claude, pricing, 1M-token output, safety features and how to get access.