AIXDAabout 2 hours ago
28.9M LLM runs on ESP32-S3
1 min read
A developer squeezed a 28.9M-parameter LLM onto an ESP32-S3. The model uses memory-mapped flash for a 25M-param embedding table and a ~560K dense core in fast RAM. It runs fully on-chip at ~9 tokens per second, telling coherent stories on the tiny microcontroller.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
