AI HOT (Curated Pool)Jun 8, 2026, 22:37 UTC+8

Xiaomi MiMo-V2.5-Pro-UltraSpeed Exceeds 1,000 Tokens/s

小米 MiMo-V2.5-Pro-UltraSpeed 突破 1，000 tokens/s，单台 8-GPGPU 节点运行 1T MoE 模型

Xiaomi MiMo and TileRT_AI released MiMo-V2.5-Pro-UltraSpeed, running a 1T MoE model above 1,000 tokens/s on a single standard 8-GPGPU node, with UltraSpeed API priced at 3x and applications open from June 8 to 23 PDT.

Why it matters: HKR-H/K/R all pass: Xiaomi MiMo gives a concrete claim of a 1T MoE exceeding 1,000 tokens/s on one 8-GPGPU node. The score stays at 80 because this is single-source and lacks task mix, precision, latency, and cost details.

Read the original ↗Export Markdown

Take

Xiaomi MiMo is flexing inference engineering, not model intelligence. The sharp claim is specific: MiMo-V2.5-Pro-UltraSpeed runs a 1T MoE above 1,000 tokens/s on a single standard 8-GPGPU node, explicitly not Cerebras or Groq hardware. That pushes against the story that fast generation needs exotic inference silicon. TileRT_AI’s presence matters; this smells like a kernel and serving-stack win more than a pure model release.

I don’t fully buy the “10x better output experience” line yet. The snippet gives a 3x UltraSpeed API price and a June 8–23 PDT application window, but no batch size, concurrency, time-to-first-token, context length, or P95 latency. For enterprise users, 1,000 tok/s is cheap theater if queueing hides the cost. If P95 stays clean under real concurrency, Xiaomi just put pressure on Groq/Cerebras-style inference positioning.

Source: AI HOT (Curated Pool) · x.com
Published: Jun 8, 2026, 22:37 UTC+8 · Jun 8, 2026