What OpenAI Announced
On October 8, 2026, OpenAI announced "Ultrafast" mode for GPT-6.1 Sol in the OpenAI Developer Community, with rollout across the API, Codex, and ChatGPT Work. OpenAI describes it as "near-Astra intelligence at up to 8x the speed of Sol Standard" — the same GPT-6.1 Sol model, served on infrastructure tuned purely for latency.
A New Pricing Dimension: Speed Sold Separately
The notable move is pricing. Ultrafast is the first time OpenAI charges separately for speed rather than capability: in the API, Standard runs $2/$10 per million input/output tokens, Fast $4/$20, and Ultrafast $12/$60. Ultrafast costs 1.2x the price of GPT-6 Astra and 6x the cost of Sol Standard — and supports prompts up to 272K tokens at the base rate, rising to $24/$90 above that threshold.
| Tier | Input per 1M | Output per 1M |
|---|---|---|
| Standard | $2 | $10 |
| Fast | $4 | $20 |
| Ultrafast (up to 272K tokens) | $12 | $60 |
| Ultrafast (above 272K tokens) | $24 | $90 |
What It Means for Codex and ChatGPT Work Users
In Codex and ChatGPT Work, Ultrafast is gated behind the Pro $500 plan, eligible usage-based Enterprise plans, or credit-based Edu plans. It consumes included usage at 8x and purchased credits at 6x the Standard rate — so casual users won't hit it accidentally, but teams paying for speed get a simple lever to pull.
Who Actually Needs It
OpenAI positions Ultrafast for work where speed and intelligence both matter: debugging a live outage, AI agents navigating apps in real time, and interactive experiences where every second counts. The pattern is clear — the model no longer has a single price; the price depends on how fast you need the answer.
Frequently Asked Questions
What is GPT-6.1 Sol Ultrafast mode?
It is a low-latency serving tier for OpenAI's GPT-6.1 Sol model, announced October 8, 2026. It delivers near-Astra intelligence at up to 8x the speed of Sol Standard, rolling out in the API, Codex, and ChatGPT Work.
How much does Ultrafast cost?
In the API it costs $12 per million input tokens and $60 per million output tokens (up to 272K-token prompts), which is 6x Sol Standard and 1.2x GPT-6 Astra. This is the first time OpenAI has priced speed separately from model capability.
Who can use Ultrafast in ChatGPT Work?
In Codex and ChatGPT Work, Ultrafast requires the Pro $500 plan, eligible usage-based Enterprise plans, or credit-based Edu plans. It consumes included usage at 8x and purchased credits at 6x the Standard rate.
Does Ultrafast have separate rate limits?
Yes — in the API, Ultrafast is available to all API users and runs on rate limits separate from Standard/Fast tiers, with US and EU data residency supported.
Related Prompts
Dark Seinen Anime Key Visual Prompt — Neon Rain Sword Scene (9:16)
ChatGPT
Heritage Kebaya Portrait Prompt — Javanese Traditional Look (4:5)
ChatGPT
Classroom Window Light Portrait Prompt — B&W Geometric Shadows (4:5)
ChatGPT