New branch reason: No prior New Runtime record was supplied, so this announcement begins a distinct analytical record rather than being attached to an assumed related story.
Thesis
Ultrafast makes a latency-versus-token-price tradeoff explicit for developers considering GPT-6.1 Sol in time-sensitive workflows.
Observed facts
OpenAI says GPT-6.1 Sol Ultrafast is rolling out in the API, Codex, and ChatGPT Work.
1The announcement gives API rates of $12 per million input tokens and $60 per million output tokens, and says speed is up to 8x Sol Standard.
1It describes the price as 1.2x Astra.
1In Codex and ChatGPT Work, the listed plans are Pro 500, eligible usage-based Enterprise, and credit-based Edu, with Enterprise administrator enablement required.
1
Mechanism
A faster service tier can reduce waiting time in interactive or chained agent work, but the stated per-token rates make the cost of that latency reduction dependent on both token volume and the input/output mix. The announcement frames the tier for outage debugging, agents navigating apps, and live experiences. Those are intended use cases, not evidence that every workload gains the same speed or value.
Why now
The rollout and published rates give API users a concrete basis to evaluate a new speed tier rather than treating speed as an unspecified capability.
Implications
Developers can compare the stated premium against the value of lower latency for their own workloads.
Plan eligibility and admin enablement may constrain access outside the API.
Evidence boundary
The source is OpenAI’s public announcement; it supplies company-stated speed, pricing, rollout, and eligibility details, but no independent benchmark, latency distribution, or workload-specific cost data.
Watch conditions
Independent measurements establish how Ultrafast latency varies by task and workload.
OpenAI clarifies API access conditions and reports broader rollout status.