GPT-6.1 Sol halves the price of cached input
The new model costs $2 per million input tokens and $10 per million output, while cached reads drop from $0.20 to $0.10.
Alongside Dots, OpenAI released GPT-6.1 Sol, which it says comes close to Astra-level results at a fraction of the price. Input and output prices are unchanged from GPT-6 Sol at $2 and $10 per million tokens.
The one price that moved is cached input, now $0.10 per million tokens instead of $0.20. That matters most for agents, which re-read the same long context at every step of a task.
For people who run self-hosted agents like OpenClaw with their own API key, a cheaper cache can cut the cost of long-running automations noticeably, because most of an agent's tokens are repeated context.
What it means for you
If you pay per token for an agent, check that prompt caching is on. The discount only applies to cached reads.