A reproducible method for running Gemma-4 26B mixture-of-experts model on a desktop CPU without a GPU, achieving ~124 tokens/second batched inference.
Raw Developer Origin & Technical Request
Hacker News
Jun 30, 2026
I wanted to know how fast a 26B mixture-of-experts model could run on a desktop CPU with no GPU. Got ~40 tok/s single-stream (lossless) and ~124 batched. The surprising part was the byte budget: for this model you compress the output head (32% of per-token bytes), not the experts (16%). The writeup has the bandwidth roofline and the dead-ends; the repo has the reproducible recipe. Happy to answer questions.Repo: github.com/arun-prasath2005/...
Developer Debate & Comments
Frequently Asked Questions
Market intelligence mapped to A reproducible method for running Gemma-4 26B mixture-of-experts model on a desktop CPU without a GPU, achieving ~124 tokens/second batched inference..
What problem does A reproducible method for running Gemma-4 26B mixture-of-experts model on a desktop CPU without a GPU, achieving ~124 tokens/second batched inference. solve?
How is the developer community reacting to A reproducible method for running Gemma-4 26B mixture-of-experts model on a desktop CPU without a GPU, achieving ~124 tokens/second batched inference.?
What architecture is tied to A reproducible method for running Gemma-4 26B mixture-of-experts model on a desktop CPU without a GPU, achieving ~124 tokens/second batched inference.?
Are there startups building around A reproducible method for running Gemma-4 26B mixture-of-experts model on a desktop CPU without a GPU, achieving ~124 tokens/second batched inference.?
Engagement Signals
Cross-Market Term Frequency
Quantifies the cross-market adoption of foundational terms like GPU and CPU by tracking occurrence frequency across active SaaS architectures and enterprise developer debates.
SaaS Metrics