Tuesday, November 17, 2026
AI desk
/
/
OpenAI Launches GPT-6.2 Astra and Sol for ChatGPT Work and Codex

OpenAI Launches GPT-6.2 Astra and Sol for ChatGPT Work and Codex

OpenAI launched GPT-6.2 Astra and Sol for ChatGPT Work and Codex, with new pricing, access rules, API IDs, and safety classifications.
Last updated
November 17, 2026
7 min read
Fact-checked

Photo: TechJournal

Share

Quick Answer

GPT-6.2 Astra and Sol are OpenAI’s new models for ChatGPT Work and Codex, released November 17, 2026. Astra targets the hardest projects, while Sol is intended for everyday work. Both carry Critical cybersecurity and High biological-and-chemical classifications. Organizations should enable access deliberately, select the lower-cost model when it fits, and treat higher-risk requests as governed work.

Key Takeaways

  • GPT-6.2 Astra is OpenAI’s model for harder projects, while GPT-6.2 Sol is positioned for everyday work.
  • Both models are available in ChatGPT Work and Codex for eligible Plus, Pro, Enterprise, and Edu users.
  • GPT-6.2 Astra costs $8 per million input tokens, while GPT-6.2 Sol costs $2 per million input tokens.
  • Tool calling requires the Responses API, even though both models also support Chat Completions.
  • OpenAI classifies both models at its Critical cybersecurity threshold and High biological-and-chemical threshold.

What are GPT-6.2 Astra and Sol?

GPT-6.2 Astra and Sol are two new OpenAI models designed for different levels of work complexity. OpenAI introduced both models on November 17, 2026, describing Astra as the option for its hardest projects and Sol as the model for everyday work. OpenAI’s product announcement establishes the two-model approach and their availability in ChatGPT Work and Codex.

GPT-6.2 Astra is the higher-priced model, which indicates that organizations should reserve it for work where the added capability is worth the additional token cost. GPT-6.2 Sol is the less expensive option and is positioned for routine workflows that still need an advanced OpenAI model. The practical distinction is not a simple quality ranking because OpenAI assigns each model a different intended role.

OpenAI’s release adds another model choice for businesses already evaluating advanced AI systems for coding, planning, analysis, and other work tasks. Readers comparing the release with prior OpenAI security-focused models can also review an earlier OpenAI cyber model, which covered a separate model release with a cybersecurity emphasis.

Who can use GPT-6.2 Astra and Sol?

GPT-6.2 Astra and Sol are available to Plus, Pro, Enterprise, and Edu users through ChatGPT Work and Codex. Enterprise and Edu users need an administrator to enable access, which gives organizations a control point before the models become broadly available to employees and students.

GPT-6.2 access also differs by Pro subscription level for the Ultrafast option. OpenAI says Pro 500 subscribers can use Ultrafast with both Astra and Sol, while Pro 200 subscribers receive Ultrafast access starting with GPT-6.2 Sol. The difference matters because model access and response modes can vary even within paid plans.

Organizations should review administrator settings before turning on either model for a large group. GPT-6.2 Astra and Sol are not simply consumer chatbot options in this release, because their availability includes workplace and educational environments where access decisions can affect data handling, internal policies, and employee workflows.

How do GPT-6.2 Astra and Sol compare on API pricing?

GPT-6.2 Astra costs more than GPT-6.2 Sol across input, cached-input, and output tokens. Astra costs $8 per million input tokens, $0.40 per million cached-input tokens, and $40 per million output tokens. Sol costs $2, $0.10, and $10 for the same respective categories.

ModelInput priceCached-input priceOutput priceOpenAI positioning
GPT-6.2 Astra$8 per million tokens$0.40 per million tokens$40 per million tokensHardest projects
GPT-6.2 Sol$2 per million tokens$0.10 per million tokens$10 per million tokensEveryday work

GPT-6.2 Astra costs 4 times as much as Sol in each listed API token category. That pricing difference gives developers a direct reason to route routine workloads to Sol and reserve Astra for tasks that justify higher spending. OpenAI also says Astra’s input and output prices are 20% lower than GPT-6 Astra’s, with a 60% cached-input discount.

GPT-6.2 API cost still depends on how many tokens an application sends and receives, so the per-million-token prices are not a flat monthly bill. Teams should measure actual input, cached-input, and output usage before moving a production workflow to Astra. Businesses that use AI for spending analysis should also consider the sensitivity of the information involved, particularly as AI services expand into personal finance insights.

Which APIs support GPT-6.2 Astra and Sol?

GPT-6.2 Astra and Sol support the Responses API and Chat Completions, but tool calling requires the Responses API. OpenAI lists the model IDs as gpt-6.2-astra and gpt-6.2-sol in its API changelog and model availability notice.

Tool calling is the important limitation for developers deciding whether an existing integration can move directly to the new models. An application can use Chat Completions with either model, but a workflow that needs the model to call tools must use the Responses API instead. The practical response is to review an application’s current API architecture before changing a production model ID.

Developers should also test the workflow after switching models rather than assuming that an existing prompt or integration behaves identically. GPT-6.2 Astra and Sol have distinct roles, and tool-enabled workflows can carry more operational risk than a standalone text prompt because they may connect model output to business systems or user actions.

Why do GPT-6.2 Astra and Sol have elevated safety classifications?

GPT-6.2 Astra and Sol meet OpenAI’s Critical cybersecurity threshold and High biological-and-chemical threshold. OpenAI also calls Astra its first externally deployed model to meet the High threshold for AI self-improvement capabilities. OpenAI’s GPT-6.2 system card describes those classifications and the company’s related safeguards.

The safety classifications matter because OpenAI is identifying these models as capable enough to require stronger handling for certain kinds of high-risk work. A classification is not a claim that every ordinary use is dangerous. Instead, it signals that the company has assessed specific capability areas, including cybersecurity and biological-and-chemical topics, at elevated thresholds.

OpenAI says it applies monitoring to external use of both models and has tightened restrictions for higher-risk cybersecurity requests. The limitation is that users should not treat model access as permission to pursue security testing or other sensitive work without authorization. Organizations should define acceptable uses, keep human review in place, and use established security processes when work touches real systems.

What should organizations do before enabling GPT-6.2 Astra and Sol?

Organizations should set access rules before enabling GPT-6.2 Astra and Sol for Enterprise or Edu users. Administrator-controlled availability provides a useful point to decide which teams need Astra, which teams can use Sol, and which requests require additional review because of cost, data sensitivity, or security implications.

  1. Review which users need access to ChatGPT Work or Codex before an administrator enables the models.
  2. Assign GPT-6.2 Sol to routine work where lower token costs fit the task.
  3. Reserve GPT-6.2 Astra for projects that genuinely need the model positioned for harder work.
  4. Test tool-enabled workflows through the Responses API before deploying them to production.
  5. Document a stop line for security-sensitive requests and escalate them to authorized security staff.

Organizations should stop and involve internal security, legal, or compliance teams when a proposed use involves sensitive systems, regulated data, cybersecurity testing, or biological-and-chemical subject matter. OpenAI’s monitoring and restrictions are part of the service’s safety approach, but they do not replace an organization’s own authorization and governance process.

Teams using AI tools for coding should also keep credentials, tokens, and production access separate from ordinary experimentation. That precaution remains important after incidents involving exposed keys and compromised extensions, including the reported Fireworks key exposure connected to a trojanized coding extension.

Should developers choose GPT-6.2 Astra or GPT-6.2 Sol?

Developers should choose GPT-6.2 Sol for everyday work and consider GPT-6.2 Astra for harder projects where the higher price is justified. OpenAI’s positioning and API pricing make Sol the practical default for teams that need to manage recurring token costs, while Astra is the option for more demanding work.

GPT-6.2 Astra has a clear cost tradeoff. Astra costs $40 per million output tokens, compared with $10 per million output tokens for Sol, so output-heavy workflows can produce a large cost difference. Teams should test both models on representative tasks, measure token usage, and compare the quality needed for the actual business requirement.

GPT-6.2 Sol also has limits because lower pricing does not make it automatically suitable for every complex workflow. The most sensible approach is to begin with Sol, document where performance does not meet the project’s needs, and move only those cases to Astra. That approach gives organizations a reasoned model-selection process instead of treating the more expensive model as the default.

What does the GPT-6.2 release mean for ChatGPT Work and Codex users?

GPT-6.2 Astra and Sol give ChatGPT Work and Codex users two models with different cost and workload positions. The release expands the available choices for eligible paid users, while Enterprise and Edu organizations retain administrator control over whether the models are enabled.

GPT-6.2 Astra and Sol also make safety governance more relevant to everyday product decisions. Both models carry the same Critical cybersecurity and High biological-and-chemical classifications, even though OpenAI presents Astra and Sol as options for different kinds of work. Users should therefore match access to a real business purpose, not simply select a model because it is newly available.

The release arrives as AI providers add more agent-like functions and connections to workplace tools. Users should be particularly careful before allowing AI systems to take actions, access accounts, or handle payments, as seen in the growing market for AI agent payment approvals. Human confirmation and least-privilege access remain important controls for any model-connected workflow.

FAQ

What is GPT-6.2 Astra?

GPT-6.2 Astra is OpenAI’s new model for its hardest projects. OpenAI made Astra available in ChatGPT Work, Codex, and the API on November 17, 2026, with higher token pricing than GPT-6.2 Sol.

What is GPT-6.2 Sol?

GPT-6.2 Sol is OpenAI’s model for everyday work. GPT-6.2 Sol has lower API pricing than Astra and supports the Responses API and Chat Completions.

Can Enterprise users access GPT-6.2 Astra and Sol?

Enterprise users can access GPT-6.2 Astra and Sol after an administrator enables the models. Edu users also require administrator enablement before access becomes available.

Do GPT-6.2 Astra and Sol support tool calling?

GPT-6.2 Astra and Sol support tool calling through the Responses API. Chat Completions supports both models, but OpenAI requires the Responses API for tool calling.

Are GPT-6.2 Astra and Sol safe for cybersecurity work?

GPT-6.2 Astra and Sol have Critical cybersecurity classifications under OpenAI’s framework, and OpenAI says it applies monitoring and tighter restrictions to higher-risk cybersecurity requests. Users should only conduct authorized security work and should escalate sensitive requests to qualified security staff.

Share this guide
Facebook
X
LinkedIn
Written by
AI Business & Policy Desk Ashik Ahmed is a technology editor covering the business and policy side of artificial intelligence, including the companies, deals, regulation, and competition shaping the industry. He has a background in tech & data analysis, sales, and business strategy. His reporting focuses on what major AI developments actually mean for businesses and everyday users.

In this article

The AI Brief

Guides like this, every Friday.

One email. No hype cycle.

Essential guides