Qwen3.8-Flash-Next on a single Intel Arc B70: 2.5K prefill, 47 tok/s
WizardlyBump17 · reddit · 2026-11-20
A Reddit user ran Qwen3.8-Flash-Next (IQ3XXS) on a single 32GB Intel Arc B70 via the Strata SYCL stack: 2529 t/s prefill, 47 t/s generation, and 250K-token prompts handled at 2184 t/s prefill. The post includes a full podman launch command with key params (prefill=30000, spec=4, expert-cache=20000, kv-resident=32768) and the complete hardware setup (Ryzen 7 5700X3D, B580 for display, 48GB RAM, Ubuntu 25.10).
More from Infra
- Local inference users want 512GB RAM and 1M context, says current boxes fall short — AIFlow_ML · 2026-11-20
- BitDecoding-FP4: native FP4 KV cache decoding via algorithm-system co-design — kalyan_kpl · 2026-11-20
- Prefill beats decode: SGLang wins agent runs on DGX Spark despite slower tok/s — AIFlow_ML · 2026-11-20
- Nvidia supplier commitments jump from $87B to $203B as forecasts miss low — tengyanAI · 2026-11-20
- SQLite is ~35% faster than the filesystem for 10KB blobs — here's why the syscall overhead adds up — arpit_bhayani · 2026-11-20
- NVIDIA's NarrowSpec: Halving Draft Hidden Width Still Speeds Up Speculative Decoding 1.44x — nvidia · 2026-11-20