AgentPlays
OpenAI's Ultrafast mode: the fastest API tier for GPT-6 Astra and GPT-6.1 Sol
Set service_tier to "ultrafast" with gpt-6-astra or gpt-6.1-sol in the Responses API. It is open to all API users, costs more than the Standard and Fast tiers, and has its own rate limits. OpenAI recommends a WebSocket connection for agent apps that make many tool calls in a row. On OpenAI's pricing page, Ultrafast is $12 in / $60 out per 1M tokens for GPT-6.1 Sol and $60 / $300 for GPT-6 Astra (prompts up to 272K tokens). GPT-6.1 Sol Ultrafast supports US and EU data residency; Astra supports US only. OpenAI's changelog dates Sol support to Oct 8 and Astra to Sep 29.
How you’d use it: Speed up an agent's tool-call loop when latency matters more than cost.