OpenAI says Ultrafast cut incident investigations to minutes
OpenAI said Ultrafast helped its engineers reduce incident investigations from hours to minutes, framing the lower-latency processing tier as a tool for faster security analysis, code refactoring and day-to-day workflows. The company’s account, published on 2026-10-26, describes the internal use case as part of a broader effort to speed up engineering work rather than a standalone incident-response product.
Evidence and context
OpenAI’s source material says engineers use Codex Ultrafast to investigate incidents in minutes and to support security analysis, refactoring and other routine tasks. Background material supplied with the announcement notes that OpenAI currently describes Ultrafast as a lower-latency model processing tier, not a separate product named "Codex Ultrafast." It also says current documentation lists GPT-6 Astra and GPT-6.1 Sol support in Codex and ChatGPT Work.
The wider context suggests the incident-response claim should be read as an internal workflow improvement, not a quantified universal benchmark. OpenAI’s earlier 2026-08-13 preview of Ultrafast described internal incident-response workflows that read application logs, analyze traces and recent code changes, synthesize engineer conversations, identify follow-up checks, and help prepare or validate fixes; the company also said engineers remain responsible for judgment and deployment. That preview’s "up to 14 times" speed and 750 output tokens per second applied specifically to GPT-5.6 Sol and should not be generalized to every current model.
Additional background supplied for 2026-10-26 says GPT-6 Astra Ultrafast can generate tokens up to eight times faster than Astra Standard, but that is a token-generation claim rather than a verified eightfold reduction in incident-resolution time. It also says current Ultrafast access is available on Pro 500 and eligible Enterprise and Edu plans, with workspace permissions required for Enterprise access. The supplied material does not provide an independently verified before-and-after timing study for incident handling, so the reported speed-up remains OpenAI’s internal account.