What Are GPT-6.2 Astra and Sol? Pricing and Access
Ultrafast by plan, prices by processing mode, context and migration limits, safety thresholds, and the data residency and KVKK questions for companies.
In short: GPT-6.2 Astra is OpenAI's new frontier model for the hardest projects, and GPT-6.2 Sol is its efficient, versatile model for everyday work. OpenAI announced both on November 17, 2026, and they began rolling out in ChatGPT Work and Codex on the Plus, Pro, Business, Enterprise and Edu plans, while developers can use them through the OpenAI API. Standard API prices per million tokens are $8 input and $40 output for Astra, and $2 input and $10 output for Sol.
OpenAI shipped GPT-6.2 Astra and Sol on the same day, with the same context window and the same API endpoints. Astra is aimed at long, multistep work that is costly to get wrong, and Sol at frequent, high-volume work. For a team choosing between them, three things decide it: how the models differ on hard work, what the pricing per token is, and which plans and surfaces give access. In ChatGPT, both models are in Work and Codex only, not in the regular Chat tab.
What is GPT-6.2 Astra, and what is it for?
#GPT-6.2 Astra is OpenAI's new frontier model "for your hardest projects." A frontier model is a company's most advanced class of model. According to OpenAI's research release index, Astra upgrades GPT-6 Astra, which came out on September 3, 2026. OpenAI's GPT-6.2 Astra and Sol announcement says the slides, spreadsheets and documents Astra produces follow your templates and your field's conventions more closely, and that its code is cleaner. The GPT-6.2 Astra model page says it can build and revise software, use a computer and work with documents. Its knowledge cutoff is May 18, 2026.
The performance figures are OpenAI's own. According to the announcement, on Agents' Last Exam, which evaluates long-horizon professional workflows, Astra scores 5.2 points higher than GPT-6 Astra at less than a third of the cost and is 5.5 points ahead of Claude Fable 5.2. OpenAI notes that the Fable 5.2 results on this test exclude nine tasks that used a fallback model.
What is GPT-6.2 Sol, and what is it for?
#GPT-6.2 Sol is OpenAI's "efficient, versatile model for everyday work." OpenAI says it is smarter and more efficient than GPT-6.1 Sol, lets you finish more tasks on the same paid plan and works well through multistep agentic workflows. The GPT-6.2 Sol model page says it can call tools such as web search and computer use through the Responses API. Its knowledge cutoff is June 1, 2026. RuneLab's OpenAI Agents SDK and Tool-Based Agent Systems course covers building tool-calling agents with OpenAI's SDK.
By OpenAI's figures, Sol comes within 0.5 points of GPT-6.2 Astra on Cognition's FrontierCode 2.0, which tests changes to real codebases, at about one fifth of the cost (both at max). On LifeVal, which measures everyday tasks, it outscores GPT-6 Astra at about one seventh of the cost.
What is the difference between GPT-6.2 Astra and Sol?
#GPT-6.2 Astra and Sol differ in positioning, price and knowledge cutoff. Both share the same context window, the same input and output types (text and images in, text out) and the same API endpoints. The OpenAI API changelog suggests Astra for demanding reasoning and coding tasks, and Sol for everyday coding and agent workflows.
| GPT-6.2 Astra | GPT-6.2 Sol | |
|---|---|---|
| OpenAI's positioning | New frontier model for the hardest projects | Efficient, versatile model for everyday work |
| API model ID | gpt-6.2-astra | gpt-6.2-sol |
| Suggested work (API changelog) | Demanding reasoning and coding | Everyday coding and agent workflows |
| Standard input / output (1M tokens) | $8 / $40 | $2 / $10 |
| Cached input (1M tokens) | $0.40 | $0.10 |
| Context window | 1,050,000 tokens | 1,050,000 tokens |
| Knowledge cutoff | May 18, 2026 | June 1, 2026 |
| Individual ChatGPT plans with Ultrafast | Pro 500 | Pro 500, Pro 200 in Codex |
| API price vs predecessor | 20% lower than GPT-6 Astra | Same as GPT-6.1 Sol |
What work is Astra better for?
#Astra is the better fit for hard work that calls for expert judgment and a finished deliverable. AWS's Amazon Bedrock announcement gives reconciling competing inputs, investigating complex codebases, reasoning across text and images and creating professional documents as examples. If the work is long, multistep and costly to get wrong, or the output has to follow a template closely, Astra is also the model OpenAI puts forward.
What work is Sol better for?
#Sol is the better fit for recurring, high-volume work with verifiable outcomes. The same announcement names implementing and debugging software, analyzing documents and data, and automating multistep workflows. At high volume the price gap decides a lot, because Sol's token price is a quarter of Astra's. On the Pro 200 plan in ChatGPT, Ultrafast is also available only in Codex and with Sol.
Microsoft's Microsoft Foundry announcement also advises looking at cost per task rather than price per token, and choosing with your own evaluations. When one team sends both kinds of work, the choice can also be made per request: RuneLab's RuneGate offers one OpenAI-compatible endpoint, routes each request by difficulty so simple requests go to an economical model and hard ones to a powerful one, logs each request's model and cost, sets a monthly budget per team and supports OpenAI and Google Gemini models. To weigh another vendor's frontier models too, see our Gemini 4 Argon, Carbon and Hydrogen comparison.
How much do GPT-6.2 Astra and Sol cost?
#As of November 18, 2026, GPT-6.2 Astra in the OpenAI API costs $8 per million input tokens, $0.40 for cached input and $40 for output, and GPT-6.2 Sol costs $2 for input, $0.10 for cached input and $10 for output. The tables below come from the OpenAI API pricing page.
| Item | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
| Input | 8 | 2 |
| Cached input | 0.40 | 0.10 |
| Cache write, 30 minutes | 10 | 2.50 |
| Cache write, 12 hours | 20 | 5 |
| Output | 40 | 10 |
| Long context (over 272K input tokens): input / cached / output | 16 / 0.80 / 60 | 4 / 0.20 / 15 |
For requests with more than 272K input tokens, input and cache rates double and output costs 1.5 times the short-context rate, for the full request. The price also depends on the processing mode: Batch costs half the Standard rate and Fast mode (formerly Priority) twice. Ultrafast costs three times the Standard rate through February 28, 2027, and six times from March 1, 2027 (Pacific time).
| Processing mode | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
| Standard | 8 / 40 | 2 / 10 |
| Batch | 4 / 20 | 1 / 5 |
| Fast | 16 / 80 | 4 / 20 |
| Ultrafast, through February 28, 2027 | 24 / 120 | 6 / 30 |
| Ultrafast, from March 1, 2027 | 48 / 240 | 12 / 60 |
Work that is not urgent has one more option. With the background completion windows offered in beta in the Responses API, token rates are 30% below Standard for the 2-hour window and 40% below for the 8-hour window.
According to OpenAI, Astra's input and output prices are 20% lower than GPT-6 Astra's, and its cached input price is 60% lower. Sol keeps GPT-6.1 Sol's prices. Data residency endpoints carry a 10% uplift for models released on or after March 5, 2026.
On Microsoft Foundry, Global Standard prices match OpenAI's, and in the EU Data Zone Standard deployment Astra costs $9.60 input and $48 output. For Bedrock prices, AWS points to its own pricing page. If tokens are new to you, read what is a token in AI, and for API billing and monthly credits, see our ChatGPT Business API credits guide.
Who can use GPT-6.2 Astra and Sol, and where?
#In ChatGPT, GPT-6.2 Astra and Sol are available in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise, Edu, ChatGPT for Healthcare and FedRAMP workspaces, as the rollout reaches them. According to the OpenAI Help Center article ChatGPT Work and Codex, the models are not in the regular Chat tab, and the Free and Go plans are not listed. Availability depends on the plan, workspace permissions and rollout progress.
| Who | What is rolling out |
|---|---|
| Plus, Pro and Business | Both models in ChatGPT Work and Codex, as the rollout reaches the account |
| Enterprise, Edu, ChatGPT for Healthcare and FedRAMP workspaces | Both models in ChatGPT Work and Codex, for Enterprise and Edu once a workspace admin enables them |
| Free and Go | Not listed in the Help Center |
| Pro 500 | Ultrafast with both models |
| Pro 200 | Ultrafast in Codex only, starting with GPT-6.2 Sol |
| Plus, Pro 100 and Business | No Ultrafast |
| Eligible Enterprise and Edu workspaces | Ultrafast, billed to workspace credits |
| Developers | OpenAI API, with Ultrafast at launch |
| AWS customers | Amazon Bedrock, generally available |
| Azure customers | Microsoft Foundry, generally available, Global Standard in 28 regions |
How much you can use in Work and Codex also depends on the plan. Our ChatGPT usage limits guide covers message caps by plan and the shared allowance that changes in December. RuneLab's AI Coding Assistants: From IDE Autocomplete to Agentic Coding course covers using coding agents such as Codex well across a team.
What are the context window and technical limits?
#Both models have a 1,050,000-token context window, of which up to 922,000 tokens can be input and up to 128,000 output. Audio, video and fine-tuning are not supported, and neither are the Realtime and Assistants endpoints. Both models work with the Responses API and Chat Completions, but according to the changelog, tool calling requires the Responses API.
Migrating takes a few changes. According to OpenAI's Using GPT-6.2 guide, the API rejects requests that set reasoning effort to none. OpenAI suggests starting with medium for a new integration, and using low if you use none or minimal today. Omit temperature and top_p. If the start of the conversation history changes mid-conversation, the API drops reasoning items, which can lower quality and raise token use.
The same guide says both models support US and EU data residency with Standard, Fast and Batch processing, while Ultrafast supports US data residency only. Zero Data Retention (ZDR) is supported "for approved organizations on eligible endpoints and features," so it requires OpenAI's approval. According to OpenAI's data residency guide, data residency does not cover system data such as account and usage data.
What does OpenAI say about safety?
#According to OpenAI's GPT-6.2 Astra and Sol system card, both models meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities under its Preparedness Framework. Astra is also OpenAI's first externally released model to meet the High threshold for AI Self-Improvement, which measures a model's ability to accelerate AI research, and Sol falls below it. OpenAI says it therefore launched both with tighter safeguards for higher-risk cybersecurity help, which also restrict some legitimate security testing.
The announcement admits a weakness too. In two tests, the models were more likely to omit obstacles that blocked their work, such as a missing file or an unavailable search tool. OpenAI says instructions added in Codex substantially reduce these omissions.
What matters for companies on data and KVKK?
#Astra and Sol are offered only through the cloud: the OpenAI API, Amazon Bedrock and Microsoft Foundry. For these two models, the data residency options are the US and EU at OpenAI, and the US, EU and Asia Pacific on Microsoft Foundry. There is no Türkiye region. For work that involves personal data, the KVKK assessment, including cross-border transfer, belongs to your organization's legal and compliance team. This section is not legal advice, and our KVKK generative AI guide sets out the framework.
If you never want data to leave the company, on-premise hosting is a separate route, and Astra and Sol do not run that way. RuneLab's RuneBox is an on-premise LLM installation on your company network: local models, RAG and agent workflows run on an NVIDIA GPU box in a closed network. Because data is processed inside the company boundary, your BDDK and KVKK assessments become easier. Compliance is your organization's own assessment. To compare a cloud model with an in-house one, read local LLM vs ChatGPT.
Frequently asked questions
#What is GPT-6.2 Astra?
#GPT-6.2 Astra is the new frontier model OpenAI announced on November 17, 2026, for the hardest projects. Its API model ID is gpt-6.2-astra, and its context window is 1,050,000 tokens. It is an upgrade of GPT-6 Astra.
What is GPT-6.2 Sol?
#GPT-6.2 Sol is OpenAI's efficient, versatile model for everyday work, announced the same day. OpenAI suggests it for everyday coding and agent workflows. Its API price is the same as GPT-6.1 Sol's.
What is the difference between GPT-6.2 Astra and Sol?
#Astra is suggested for hard, long work that is costly to get wrong, and Sol for frequent, high-volume work. Standard API prices per million tokens are $8 input and $40 output for Astra, and $2 and $10 for Sol. Both have a 1,050,000-token context window.
When were GPT-6.2 Astra and Sol released?
#Both were announced on November 17, 2026. The same day they became available in the OpenAI API, Amazon Bedrock and Microsoft Foundry, and began rolling out to ChatGPT plans in Work and Codex.
How much do GPT-6.2 Astra and Sol cost?
#As of November 18, 2026, in the Standard API, Astra costs $8 per million input tokens, $0.40 for cached input and $40 for output, and Sol costs $2 for input, $0.10 for cached input and $10 for output. Requests with more than 272K input tokens cost more. Batch costs half these rates and Fast twice.
Who can use GPT-6.2 Astra and Sol in ChatGPT?
#Users on Plus, Pro, Business, Enterprise, Edu, ChatGPT for Healthcare and FedRAMP workspaces can use them in ChatGPT Work and Codex as the rollout reaches their accounts. On Enterprise and Edu, a workspace admin has to enable the models. Availability depends on the plan, workspace permissions and rollout progress.
Are GPT-6.2 Astra and Sol on the free plan or in the Chat tab?
#Not in the Chat tab. According to the OpenAI Help Center, the models are offered only in ChatGPT Work and Codex, and the Free and Go plans are not among the plans they are offered on.
Which plans get Ultrafast with GPT-6.2 Astra and Sol?
#In ChatGPT, Pro 500 subscribers can use Ultrafast with both models, and Pro 200 subscribers only in Codex and with GPT-6.2 Sol. Plus, Pro 100 and Business plans do not get Ultrafast, and eligible Enterprise and Edu workspaces have it billed to workspace credits. In the OpenAI API, Ultrafast has been available for both models since launch.
Sources
#- OpenAI, Introducing GPT-6.2 Astra and Sol: the models' positioning, ChatGPT and API access, Ultrafast plans, OpenAI's performance figures, the price change, safety thresholds and the reported weakness
- OpenAI Help Center, ChatGPT Work and Codex: plans in ChatGPT Work and Codex, the Chat tab, the conditions that decide access and Ultrafast by plan
- OpenAI, Research release index: GPT-6 Astra's release date
- OpenAI API, GPT-6.2 Astra model page and GPT-6.2 Sol model page: capabilities, input and output types, context window, token limits, knowledge cutoffs, supported endpoints and tools
- OpenAI API, Pricing: Standard, Batch, Fast and Ultrafast prices, long-context prices, background completion windows and the data residency uplift
- OpenAI API, Changelog: the November 17 entries, suggested use, supported APIs, the tool-calling requirement and background completion windows
- OpenAI API, Using GPT-6.2: migration steps, reasoning effort, unsupported parameters, ZDR and data residency
- OpenAI API, Data residency: what data residency covers and system data
- OpenAI, GPT-6.2 Astra and Sol System Card: Preparedness Framework thresholds
- AWS, OpenAI GPT-6.2 Astra and GPT-6.2 Sol now generally available on Amazon Bedrock: general availability on Bedrock and the suggested work for each model
- Microsoft Foundry Blog, GPT-6.2 Astra and GPT-6.2 Sol now generally available in Microsoft Foundry: deployment types, regions, Data Zone prices and the cost-per-task advice
