The announcement says the mode runs at up to eight times the speed of Sol Standard and describes its intelligence as near Astra. Those are OpenAI’s comparisons, not independently verified measurements. The API price is $12 per million input tokens and $60 per million output tokens. In Codex and ChatGPT Work, access is listed for Pro 500, eligible usage-based Enterprise, and credit-based Edu plans; Enterprise administrators must enable it.
The first implication is that builders can evaluate speed against token spend for latency-sensitive tasks instead of treating standard inference as the only option. Builders can evaluate speed against token spend for latency-sensitive tasks instead of treating standard inference as the only option. The announcement does not establish realized task-level latency, quality parity, or comparative economics; its eightfold speed and near-Astra intelligence statements remain vendor claims, without a published method in this source. What to watch: whether independent measurements and customer usage show that the speed gain holds on real workloads at the stated price. OpenAI also says the mode is available in all supported regions, including US and EU data residency; the announcement adds EU residency support for GPT-6.1 Sol Fast and GPT-6 Luna Fast.
This starts a distinct analytical record because no prior New Runtime record was supplied for this specific GPT-6.1 Sol Ultrafast rollout and its associated pricing and eligibility terms.
Thesis
A high-speed tier makes latency a selectable product and cost dimension across both API and work surfaces.
Observed facts
OpenAI says GPT-6.1 Sol Ultrafast is rolling out in the API, Codex, and ChatGPT Work, at up to eight times the speed of Sol Standard.
1The announced API price is $12 per million input tokens and $60 per million output tokens; Enterprise administrators must enable eligible access in Codex and ChatGPT Work.
1
Mechanism
OpenAI packages the mode as a named option, prices API use separately, and ties app access to specific subscription or usage plans, so teams can decide where faster responses justify the cost and access constraints.
Why now
OpenAI is presenting speed as consequential for outage debugging, agents navigating apps, and live experiences, where waiting can affect the workflow rather than merely benchmark scores.
Implications
Builders can evaluate speed against token spend for latency-sensitive tasks instead of treating standard inference as the only option.
Plan eligibility and admin enablement make access governance part of the deployment decision.
Evidence boundary
The announcement does not establish realized task-level latency, quality parity, or comparative economics; its eightfold speed and near-Astra intelligence statements remain vendor claims, without a published method in this source.
Watch conditions
whether independent measurements and customer usage show that the speed gain holds on real workloads at the stated price.