Hacker News: Front Page 2026-06-08 15:27 MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second Open original source ↗ Reindex This Article Article URL: https://mimo.xiaomi.com/blog/mimo-tilert-1000tps Comments URL: https://news.ycombinator.com/item?id=48446639 Points: 580 # Comments: 426
GateGPT: 56k tokens per second Transformer (KV cache) on FPGA at 80 MHz Hacker News: Front Page • similarity 0.480
Real-time LLM Inference on Standard GPUs: 3k tokens/s per request Hacker News: Front Page • similarity 0.429
No comments yet.