OpenAI adds Ultrafast mode to GPT-6.1 Sol: near-Astra intelligence at up to 8x the speed, $12/$60 per million tokens in the API

Why it matters

【Key points】

On Oct 8 OpenAI rolled out Ultrafast mode for GPT-6.1 Sol in the API, Codex and ChatGPT Work, pitched as "near-Astra intelligence at up to 8x faster speeds than Sol Standard" for work where seconds matter: outage debugging, agents navigating apps and live experiences.

In the API it is gpt-6.1-sol with service_tier set to ultrafast, available to all API users with rate limits separate from Standard and Fast, global processing and US and EU data residency. Pricing is $12 per million input tokens and $60 per million output, versus $2 and $10 for Sol Standard. In Codex and ChatGPT Work it is limited to Pro 500, eligible usage-based Enterprise and credit-based Edu plans, with admin opt-in for Enterprise. Ultrafast first arrived on GPT-6 Astra on Sep 29 with US residency only; OpenAI recommends WebSockets to cut network overhead across many tool calls.

【Verifiable】

Ultrafast docs: https://developers.openai.com/api/docs/guides/ultrafast-mode

OpenAI developer changelog: https://developers.openai.com/changelog

@OpenAIDevs: https://x.com/OpenAIDevs/status/2108262812489531498

【Our read】

1. Speed now has its own price: the same model, 6x the cost for up to 8x the speed. Latency becomes a tier you buy, not just a reason to switch to a smaller model.

2. With GPT-6.1 Astra shelved in late September, Sol looks like the workhorse. GPT-6 Astra already had Ultrafast; extending it to cheaper Sol shows OpenAI filling the missing-flagship gap with "near-Astra, and fast."

3. Tool-heavy agents gain most, since waits stack up per step. 8x is OpenAI's ceiling; real gains depend on the separate rate limits and on using WebSockets.