GPT-6.1 Sol lowers cached-input pricing for repeated agent work

GPT-6.1 Sol upgrades OpenAI’s mid-tier model for coding, computer use and professional work. Its standard API rates are $2 per million input tokens and $10 per million output tokens, with cached input reduced to $0.10. OpenAI reports gains over GPT-6 Sol, making repeated tasks with reusable context a useful place to evaluate the new model.

OpenAI · Sep 29, 2026

Why it matters

Agent builders can compare total task cost when many requests reuse context.

More context

Try gpt-6.1-sol against your existing task suite. It is available in the API, Codex and ChatGPT Work on listed paid plans, but not yet ordinary Chat. The announced Ultrafast option is forthcoming.

Keep in mind

Performance comparisons are OpenAI’s reported evaluations, not 1min measurements. Cache discounts only apply to eligible cached input, and benchmark gains do not guarantee gains on your own workload.

Original read