OpenAI launches GPT-6.1 Sol, halves cached-input price

OpenAI launched GPT-6.1 Sol on September 29 as an upgrade for coding, computer-use and professional-work tasks, making the model available in ChatGPT Work, Codex and its API.

For developers and workplace customers, the model is priced at $2 per million standard input tokens, $0.10 per million cached input tokens and $10 per million output tokens. The standard input and output rates are unchanged from GPT-6 Sol. The direct reduction is in cached input, which fell by half from $0.20 per million tokens. Cached input covers reused prompt context, so the lower rate applies when an application sends previously processed context again.

Developers can call the model through the API as gpt-6.1-sol. OpenAI said access began September 29 for Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex. The company said the model was not yet available in Chat.

OpenAI described GPT-6.1 Sol as nearly matching GPT-6 Astra on agentic coding, computer use and professional work at one-fifth of Astra’s standard input and output token prices. The comparison is based on OpenAI’s evaluations and has not been independently replicated in the sources reviewed for this story.

On DeepSWE v1.1, a benchmark built around long-running software-engineering tasks in real codebases, OpenAI said GPT-6.1 Sol matched GPT-6 Astra at roughly one-fifth of the cost. The company also said the new model beat GPT-6 Sol’s best result by 6.4 percentage points at a lower reasoning-effort setting and lower cost.

OpenAI reported similar results for computer use. On the offline portion of OSWorld 2.0, the company said GPT-6.1 Sol scored seven percentage points above GPT-6 Sol at maximum reasoning effort for less than half the cost per task. It said the new model came within 2.1 points of GPT-6 Astra at about one-seventh the cost per task.

At low reasoning effort, OpenAI said 7.7% of GPT-6.1 Sol responses contained a factual error, compared with 11.4% for GPT-6 Sol. The company cautioned that the evaluation used deliberately difficult conversations in which users had flagged a previous model’s error and was not representative of typical use.

OpenAI also said GPT-6.1 Sol performed better than GPT-6 Sol in its alignment evaluations and made no attempts to bypass an automated safety reviewer. Those safety results remain company claims.