llama.cpp Fork Adds Qwen3.8-Flash Support, Hits 44 tks on RTX 5090 at Q4
giveen · reddit · 2026-08-28
Developer giveen submitted PR #324 to TheTom/llama-cpp-turboquant, porting qwen4exp (Qwen3.8-Flash-Next) support from upstream vLLM PR #27742. Benchmarked on an RTX 5090 at Q4 quantization with --moe-cache auto enabled, it reaches roughly 44 tokens/s.
More from Infra
- MiniCPM-V 4.7 runs a 35B MoE multimodal model on phones with just 3GB memory — nevrekaraishwa2 · 2026-10-15
- Nvidia Is Now Worth More Than the Entire Indian Stock Market — theteknosaur · 2026-10-15
- NVIDIA joins EU's AI in Science Enablers, offering compute, training and cloud resources — nordicinst · 2026-10-15
- Open source's hardware trap: data centers crowd out consumer GPUs, local LLMs become a privilege — Dante_77A · 2026-10-15
- NVIDIA backs EU AI in Science push as 35 supercomputers to serve 3M researchers — NVIDIA Blog · 2026-10-15
- Jensen Huang: 2027 AI bottleneck is 35GW of power, revenue justifies the capex — BG2 Pod · 2026-10-15