OpenAI shipped GPT-6.1 Sol alongside Dots and says it gets close to Astra-level results for a fraction of the price. Standard pricing matches GPT-6 Sol, with input at $2 and output at $10 per million tokens.
The change is in cached input, which falls from $0.20 to $0.10 per million tokens. Agents benefit most, because they re-read the same long context at each step of a task.
Self-hosted agents such as OpenClaw that run on your own API key can see a noticeable drop in cost for long automations. Most of the tokens such an agent consumes are repeated context.