On October 8, 2026, OpenAI added Ultrafast for GPT-6.1 Sol in the Responses API. Developers can select gpt-6.1-sol with service_tier set to ultrafast.
Understand the latency target
The tier reduces time between generated output tokens. A workflow dominated by searches, database calls, or external tools may still spend most of its time elsewhere. Measure response onset, generation duration, and end-to-end completion separately.
Separate API and subscription access
The announcement makes the tier available to API users subject to rate limits, with global processing and US and EU data residency. This differs from Astra Ultrafast in the personal Pro 500 subscription. API requests follow their own service-tier pricing.
Compare equivalent work
Use tasks with similar inputs, outputs, and reasoning settings to compare speed and total cost. A faster service tier does not remove the need to validate answers and tool results. Decide whether reduced waiting has sufficient value for the workload.
