OpenAI is previewing Ultrafast, a new API tier powered by Cerebras, according to a Techmeme item summarizing Zac Hall at 9to5Mac. The report says the tier runs GPT-5.6 Sol and is positioned as a faster way to use OpenAI’s most capable GPT-5.6 model through the API. The headline figures are latency and output speed. The report says Ultrafast can run GPT-5.6 Sol up to 14× faster and generate up to 750 output tokens per second. Those are “up to” claims, so they should be treated as peak or best-case figures until OpenAI provides broader detail on workloads, rate limits, and availability. The provided material does not specify pricing, launch timing beyond preview status, customer eligibility, regions, or benchmark methodology. It also does not say whether Ultrafast changes model behavior, context limits, or API features; the supported claim is narrower: OpenAI is previewing a Cerebras-powered tier focused on faster inference for GPT-5.6 Sol. Who benefits: OpenAI API customers that need lower-latency generation could benefit if Ultrafast becomes broadly available on workable terms. Cerebras also gains visibility as the compute provider named in the preview. Who's exposed: Inference providers competing on speed and throughput may face more pressure if OpenAI turns the preview into a generally available tier. The provided material is too limited to assess pricing or margin impact.