Custom CPU engine for Qwen3.5 0.8B beats llama.cpp with 2.9x faster prefill
Danmoreng · reddit · 2026-09-08
Reddit user Danmoreng vibecoded a small C++ inference engine and a custom 4-bit quant format (H128/Q4-G32-DOT4, with activation-based calibration and blockwise error compensation) to run Qwen3.5 0.8B on CPU — his use case is local dictation cleanup while the GPU is busy.
- 425 MB weight payload, 71 MB smaller than Unsloth's mixed-precision Q40, with similar perplexity and KL divergence
- On a Ryzen 9 9955HX3D (8 V-Cache cores) vs llama.cpp Q40: 2.9x prefill, 1.3x single-request decode, 1.7x batch-16 throughput
- vs ikllama.cpp on the same GGUF: 18% faster single-request decode, 53% higher batch-16 throughput
- Open source at github.com/Danmoreng/qwen35-cpu
More from Infra
- HBM demand shifts: low-stack inference plans shelved, Rubin to go 12-high only — BenBajarin · 2026-10-26
- CircuitHub unveils The Grid, a factory-scale robot making PCBs in the US twice as fast as Shenzhen — garrytan · 2026-10-26
- Ubicloud publishes "Fixing Tool-Call Accuracy in vLLM" — petrusenko_max · 2026-10-26
- Meta approves nonvoting Class C shares as 2026 capex guidance hits $135B–$145B — thione · 2026-10-26
- Unsloth Ships MiniMax-M3.1-Flash Quants: 2-bit Retains 91.5% Accuracy, Runs in 150GB RAM — danielhanchen · 2026-10-26
- Engineer with thousands of fabricated devices pours cold water on photonic crystal modulator yields — jwt0625 · 2026-10-26