OpenAI released GPT-6.1 Sol on September 29. The company says it approaches its more powerful GPT-6 Astra on selected coding and work tasks, at one-fifth the standard price per token. Sol is available through the API and on ChatGPT Plus, Pro, Business, Enterprise and Edu plans.

The lower rate can make repeated API tasks cheaper, but it does not guarantee that any particular job will cost one-fifth as much. The total depends on how many input and output tokens the task uses.

What the price difference means

OpenAI lists these standard API prices per million tokens:

  • GPT-6.1 Sol: $2 for input and $10 for output.
  • GPT-6 Astra: $10 for input and $50 for output.

Those rates make Sol one-fifth the price per token. A task that uses different amounts of tokens on each model may produce a different overall saving.

What OpenAI’s benchmark shows

OpenAI describes Sol as a major upgrade to GPT-6 Sol, with improvements in agentic coding, computer use and professional work. The company says Sol matched Astra on its DeepSWE v1.1 coding evaluation at roughly one-fifth the cost per task.

That is a company-reported result, not proof that the models perform equally across software projects. DeepSWE v1.1 tests complex engineering tasks in real codebases. OpenAI advises customers to compare the models on their own work before choosing between Sol’s lower price and Astra’s greater capabilities.

Who can use Sol

Sol is available now, not only in preview. API customers can use it for coding and computer-use workflows. ChatGPT subscribers on the listed paid plans can access it through the service. OpenAI’s DevDay recap does not list the Free plan as eligible.

For teams running repeated or token-heavy tasks, the lower rates could help stretch an API budget. The practical next step is to test both models on the same work: actual performance and token use can vary by task.