The Pulse
OpenAI Adds a 6×-Priced Ultrafast Tier for GPT-6.1 Sol
OpenAI added Ultrafast mode for GPT-6.1 Sol in the Responses API on October 8, offering shorter intervals between generated output tokens at six times Standard pricing. Prompts exceeding 272K input tokens have separate input, cache and outp
AI.info Team · Written with AI ·
Six times the price for shorter waits
OpenAI added Ultrafast mode for GPT-6.1 Sol to the Responses API on October 8. Developers request it with service_tier: "ultrafast" while keeping the model name gpt-6.1-sol. The changelog says the mode is intended to reduce the time between generated output tokens. It is available to API users subject to rate limits, with global processing and US and EU data-residency options.
For prompts with up to 272K input tokens, GPT-6.1 Sol’s Standard rates are $2 per million input tokens, $0.10 per million cached input tokens, $2.50 per million cache-write tokens and $10 per million output tokens. The model page says Ultrafast prices are six times Standard. On that basis, the corresponding Ultrafast rates are $12 per million input tokens, $0.60 for cached input, $15 for cache writes and $60 for output. Costs vary with token use, not simply with the number of requests.
Those figures do not cover every request. For prompts exceeding 272K input tokens, OpenAI lists rates of twice the input and cache rates and 1.5 times the output rate for the full request. The model page also says regional processing adds a 10% premium where available. Developers evaluating costs should account for these conditions as well as the base Ultrafast multiplier.
Available across API accounts, within rate limits
OpenAI says Ultrafast is available to all API users, subject to rate limits. Availability does not mean unlimited throughput: account rate limits still apply. GPT-6.1 Sol supports US and EU data residency, including with Fast and Ultrafast modes, and the changelog describes global processing as an option for the new tier.
The model documentation directs developers to use the Responses API for tool calling; Chat Completions is supported without tool calling. The Ultrafast instructions specify a Responses API setting, so developers should not assume the same option applies to every endpoint. The model page also lists other supported endpoints, but that does not change the documented instructions for activating Ultrafast.
Token cadence, not a published latency guarantee
OpenAI describes the benefit narrowly: less time between generated output tokens. The rollout note does not provide a measured tokens-per-second rate, a first-token latency figure or a workload-specific speed comparison. The published terms establish the mode’s intended effect and pricing, but do not say how much faster a particular application will run.
That distinction matters for developers deciding whether to pay the premium. A faster stream can be useful when a person or a downstream process is waiting for a long response, but the published information does not quantify the gain for a given prompt or task. Teams will need to measure their own latency and total token costs against Standard, including the adjustments for long prompts and regional processing if those apply.
What developers can measure
GPT-6.1 Sol is described as a model for complex coding, computer use and professional work. Its model page lists a 1,050,000-token context window and a maximum output of 128,000 tokens. The Ultrafast option keeps the same model identifier and adds a service-tier choice; the announcement does not describe a separate model version or a different quality level.
For developers, the published tradeoff is a sixfold Standard price for reduced spacing between output tokens, with rate limits still in effect. The listed token prices provide a basis for estimating usage, while actual application latency and total cost depend on the workload and the pricing conditions that apply to each request.