OpenAI introduced GPT-6.2 Astra and Sol on November 17, with API model IDs gpt-6.2-astra and gpt-6.2-sol. The release is relevant to people choosing a model for work, but the available measurements do not support one blanket verdict about which model is best for every task.

Access and API cost

OpenAI lists Astra at US$8 per million input tokens, US$0.40 per million cached input tokens and US$40 per million output tokens. Sol is US$2, US$0.10 and US$10 respectively. These are the company’s standard API prices checked on November 19, not an estimate of the cost of a complete task. Usage, output length and caching will change the bill.

OpenAI says the models are available through its API and to eligible ChatGPT Work and Codex users. Its announcement names Plus, Pro, Enterprise and Edu plans; Enterprise and Edu access may require an administrator to enable the models. Availability for an individual account should be checked in that account.

The benchmark result depends on the task

In Artificial Analysis’s Intelligence Index v5, Astra ranked second behind Claude Fable 5.2 and Sol scored 57. Yet its Coding Agent Index, using a Codex harness at maximum effort, found Astra flat against GPT-6 Astra and Sol three points below GPT-6.1 Sol. These are measurements by Artificial Analysis, independent of OpenAI, within its own benchmark setup. They do not show how either model will perform in every coding workflow.

For a purchasing or routing decision, compare the task you actually run, the effort setting, quality requirements and total task cost. OpenAI’s product claims and Artificial Analysis’s scores are different kinds of evidence; neither replaces a test on your own workload.