GPT-6.1 Sol Ultrafast: up to 8x faster, at 6x the price

OpenAI has rolled out Ultrafast, a speed tier for GPT-6.1 Sol, across the API, Codex and ChatGPT Work. It's the same model with faster serving, not a new one. OpenAI says generation is up to 8x faster than Sol Standard in Codex and up to 6x in the API. Those are maximums, and nobody has independently measured them yet.

In the API it costs $12 per million input tokens and $60 per million output tokens. That's 6x Sol's standard $2/$10, and 1.2x the price of the bigger GPT-6 Astra. Sol is pitched as near-Astra quality, so Ultrafast gives you roughly Astra-level results at a faster speed and a somewhat higher price. A run with 1M input and 1M output tokens costs $72, against $12 on standard.

The API is open to everyone. In Codex and ChatGPT Work it's limited to Pro 500, usage-based Enterprise (admins must switch it on) and credit-based Edu plans. US and EU data residency are supported. OpenAI recommends persistent WebSocket connections for agents that chain tool calls, because network overhead can eat the latency gain.

Why it matters: for agentic coding loops, latency is the bottleneck, and OpenAI is now selling speed as a separately priced product.