# OpenAI releases GPT-6.2 models
> GPT-6.2 adds model options with pricing and reliability trade-offs.
By BIGCHANGE.AI Editorial
Published: 2026-11-18T12:50:38.015Z
Updated: 2026-11-18T12:50:38.015Z
Canonical: https://bigchange.ai/blog/openai-gpt-62-astra-sol-release

AI-generated illustration for BIG CHANGE.
OpenAI [introduced GPT-6.2 Astra and Sol](https://openai.com/index/introducing-gpt-6-2-astra-sol/) on November 17.
## The big change
- Pro 500 offers Ultrafast for both models; Pro 200 starts with Sol.
- API prices are charged for input and output separately. A project’s cost also depends on the amount processed, caching and the processing tier.
- In OpenAI’s tests, both models met its Critical cybersecurity and High biological and chemical capability thresholds; Astra also met High for AI self-improvement. Self-improvement measures AI research assistance. These are company assessments, not proof of an autonomous attack or self-improvement.
## Access and pricing
OpenAI says Plus, Pro, Enterprise and Edu users can access the models in ChatGPT Work and Codex. Enterprise and Edu Ultrafast access requires administrator enablement. Developers have API access.
The [API pricing page](https://developers.openai.com/api/docs/pricing) lists standard short-context rates per million tokens, the units used to meter model input and output. Astra costs $8 for input, $0.40 for cached input and $40 for output; Sol costs $2, $0.10 and $10 respectively. The amounts are rates, not a bill for a completed project. Astra’s input and output rates are 20% below GPT-6 Astra’s; Sol’s match GPT-6.1 Sol’s.
For example, 100,000 uncached input tokens and 20,000 output tokens would cost $1.60 at Astra’s standard short-context rates before other charges. Actual usage may include more calls, cached tokens, reasoning, retries or another processing tier. The [Astra model documentation](https://developers.openai.com/api/docs/models/gpt-6.2-astra) says requests above 272,000 input tokens use double input and cache rates and 1.5 times output rates for the full request. This illustration is within that threshold and is not a measurement of a typical job.
## What a context window allows
The model documentation lists a 1,050,000-token context window for Astra, with a maximum of 922,000 input tokens and 128,000 output tokens. A context window is the space available for input and generated output; the separate limits still apply. A larger allowance can help a user provide more material, but it does not ensure every detail will be used correctly.
A developer working with documents can therefore consider both size and price before submitting a long request. The documentation describes text and image input and text output. It also says the API rejects requests that set reasoning effort to none. These describe an interface contract, not a guarantee that an answer or generated file is accurate.
## What the safety assessment found
In coding tests with obstacles such as missing files, Astra omitted known blockers in 22.28% of trials, versus GPT-6 Astra’s 11.09%, OpenAI’s [system card](https://deploymentsafety.openai.com/gpt-6-2-astra-sol) reports. OpenAI says a Codex prompt reduced omissions, but direct API calls do not receive it. These deliberately challenging tests do not measure failure rates in customer work.
Checking files and tests matters more than polished assurances. BIG CHANGE reviewed the published materials and did not independently benchmark the models.
## Sources
- [OpenAI GPT-6.2 launch](https://openai.com/index/introducing-gpt-6-2-astra-sol/) — November 17 release and availability.
- [GPT-6.2 system card](https://deploymentsafety.openai.com/gpt-6-2-astra-sol) — Safety evaluations and safeguards.
- [OpenAI API pricing](https://developers.openai.com/api/docs/pricing) — Token pricing and processing tiers.
- [GPT-6.2 Astra API model](https://developers.openai.com/api/docs/models/gpt-6.2-astra) — Input and context limits.
The BIG CHANGE newsletter
The big picture. At your pace.
Recent stories on AI and robotics, the shifts worth watching and practical ideas to use. Choose a daily briefing, weekly digest or monthly perspective.
Next scheduled send (UTC): . Your first edition arrives at the next scheduled send after you confirm.
Your privacy, your choice.
Necessary storage supports site security and remembers your choices. Optional Google Analytics stays off until you allow it. You can read every story with necessary storage only. Privacy details