Qwen3.8-Flash-Next · Strata · 2× Tesla V100 32GB PCIe (no NVLink)
English | 简体中文 | 日本語 · ← All recipes
Source level: ✅ Measured · Date: 2026-10-09 · Numbers come from the source. This repo did not measure them again.
Hardware
- 2× Tesla V100 32GB PCIe (no NVLink) (Used datacenter GPU)
- Reference cost: Not published. See the source.
Model
- Qwen3.8-Flash-Next ·
q3 xs (Strata tutorial) - Download: Not published. See the source.
Engine
- Strata — install: https://github.com/Niko1221/Strata
Launch command
Command not published. See the source.
Key parameters
- Context: —
- Concurrency: single stream (Strata supports single stream only)
- Tags:
3bit,MoE,multi-GPU
Expected speed
- Up to 120 tok/s decode, 2,000 tok/s prefill (llama.cpp 30 / 120) (✅ Measured)
Caveats
- Speed changes with drivers, engine version, context length, and prompt length.
Source & attribution
- https://x.com/rewin123/status/2108669905616736529 — @rewin123
- Credit to the original author. This page only summarizes and links. Data under CC BY 4.0.
Source file: recipes/x-007.md