gemma4-26b-a4b

Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook

Free-first, model-adaptive coding agent runtime that makes local small LLMs work reliably with harness, checkpointing, verification, cache-aware context, and free-model fallback.

Zero-training brand exposure auditor — detect any logo in any broadcast using Gemma 4 vision on Apple Silicon

Configuration layer and prompt engineering toolkit for Gemma 4 26B-A4B. Optimized for Google AI Studio.

chatdsl for repo evaluation - uses chatybot. ( pip install chatybot )