88Z AI, 9 June — Xiaomi, in collaboration with TileRT, has launched the MiMo-V2.5-Pro UltraSpeed mode, achieving a generation speed of over 1000 tokens/s for trillion-parameter models on general-purpose GPUs. Priced at three times the standard version, it offers approximately tenfold faster output and is available exclusively via API. Access is limited and application-based, prioritizing enterprises and professional developers.
At a glance
- Trillion-parameter model generation speed on general GPUs exceeds 1000 tokens/s for the first time
- Output speed increases by approximately 10x, available via API application for enterprises and developers only
The update
Xiaomi has partnered with TileRT to introduce the MiMo-V2.5-Pro UltraSpeed mode, marking a significant breakthrough on general-purpose GPU hardware by enabling trillion-parameter models to achieve generation speeds exceeding 1000 tokens per second for the first time. This high-performance mode is priced at three times the cost of the standard version, offering an approximate tenfold increase in output speed to significantly enhance processing efficiency. Currently, this feature is available exclusively through an API interface, with no other access methods provided. To ensure effective resource allocation, the UltraSpeed mode is being released on a limited-time basis requiring an application process. Enterprises and professional developers will be given priority during the review phase to meet the demands of high-load and professional scenarios.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- IT之家Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.