AIXDAabout 2 hours ago
Old Phone Runs Local LLM Server
1 min read
A Poco M6 Pro runs llama.cpp with Gemma 4 E2B, achieving 5-6 tokens per second for OCR, document summarization, and email proofreading. The setup uses Termux and parallel compilation to speed up build time. Qwen3.6-35B-A3B remains superior for coding tasks at 24+ tokens per second.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
