On October 8, 2026, OpenAI began offering Ultrafast mode, which runs GPT-6.1 Sol, the model it announced at DevDay in September, up to eight times faster than usual. Two days earlier, on October 6, it made the Decisions API, which quickly handles decisions made inside apps, available in public beta to all developers. Both are developer-focused announcements that put speed front and center.
What is Ultrafast mode?
Ultrafast is a mode that runs GPT-6.1 Sol up to eight times faster than the standard setting. OpenAI describes it as intended for work that needs quick responses while keeping intelligence close to that of the higher-end GPT-6 Astra model. It gives the following examples of suitable use cases.
- Investigating the causes of system outages
- Agents operating apps
- Live experiences where every second counts
Replies to the announcement post explain that it also supports voice use and the ability to operate a computer (computer use).
Availability and pricing
Ultrafast is available in the API in all supported regions. In Codex and ChatGPT Work, it is available to eligible users on the Pro 500 plan, pay-as-you-go Enterprise and credit-based Edu. In Enterprise, an administrator must enable it. OpenAI's documentation also states that Ultrafast is available for GPT-6 Astra.
API pricing is $12 per million input tokens and $60 per million output tokens. According to OpenAI's API documentation, this is six times the price of standard Sol ($2 input, $10 output) and about 1.2 times that of GPT-6 Astra. Since standard Sol costs one-fifth as much as Astra, Ultrafast can be described as an option that raises costs substantially in exchange for speed. Data residency can be set to the US or the EU, and EU data residency is now also available for GPT-6.1 Sol Fast and GPT-6 Luna Fast.
What is the Decisions API?
The Decisions API takes text or images and returns answers in a fixed format to questions the developer has defined in advance. It is designed for small decisions inside apps, such as choosing which model or tool to use, classifying inquiries and deciding the next action to take. There are three answer formats.
- predicates: Returns the probability that a statement is true
- choices: Picks one option from a defined set and returns it with a confidence level
- scores: Rates according to defined levels
It uses GPT-6 Luna, and OpenAI says it can make decisions up to 10 times faster than using GPT-6 Luna through the standard Responses API. Pricing is $0.10 per million input tokens, with no charges for output or caching (surcharges apply for region-specific processing and long inputs). With this public beta, anyone can now use it, and OpenAI expects to make it generally available within a few weeks.
CoAI's take
These two announcements are strongly geared toward AI agents doing real work. A pattern is emerging of using models by role: leaving complex work to Astra or Sol, using Ultrafast for urgent tasks, and using the cheap, fast Decisions API for small decisions. However, because Ultrafast costs more than Astra, the realistic approach is not to use it for everything but to limit it to situations where speed truly matters. These are not changes for general ChatGPT users but announcements mainly for developers and businesses.
