MTP released for Qwen3.8-Flash-Next GGUF, promising big local TPS gains
vini542reddit · reddit · 2026-09-01
MTP (Multi-Token Prediction) support has been released for the GGUF quantization of Qwen3.8-Flash-Next. The poster expects it to significantly boost local tokens-per-second and is eager to test it.
What remains is for more llama.cpp optimizations to be merged so local users can benefit.
Related event: Qwen3.8-Flash Gets MTP Support for Up to 1.7x Local Speedup(2 posts)→
More from Infra
- SGLang Summit sponsors grow from 11 to 36 in one week; Lip-Bu Tan, Lilian Weng to speak — BanghuaZ · 2026-10-18
- Yacine jokes his datacenter KPI is gigawatt-hours: get paged if not burning enough power — yacineMTB · 2026-10-18
- 1.68x inference speedup drops, directly accelerating GRPO post-training — yacineMTB · 2026-10-18
- OpenRouter is only a fraction of the total addressable market, poster argues — sloppenheimer · 2026-10-18
- Developer builds change alerts for LLM prices and retirements across Bedrock, Azure and Vertex — jsandhu0780 · 2026-10-18
- 12.3M tokens in 27 minutes for $0.14: agent session with 99.6% cache hit — teortaxesTex · 2026-10-18