Gemma 4 E4B IT Q4_0 — Mesh-LLM / Skippy Layer Package

This repository contains a native Mesh-LLM / Skippy layer-package generated from Google's official Gemma 4 E4B instruction-tuned Q4_0 GGUF distribution.

Source model:

• Repository: google/gemma-4-E4B-it-qat-q4_0-gguf
• Revision: 4b4a2c1d584be7264f87aac328a1bc739ce81b6c
• Primary source file: gemma-4-E4B_q4_0-it.gguf
• Source SHA-256: 676c35070db6dbe52f93e9c864ee0fba4eddea94b9c875d9cb10daff453fbaee
• Official projector: gemma-4-E4B-it-mmproj.gguf
• Projector SHA-256: 7498a37cb619e55f2fcf87eb931f56e99389ed6d432e4c5c66110694c0d65578

Package characteristics:

• Schema: layer-package version 1
• Skippy ABI: 0.1.24
• Transformer layers: 42
• Activation width: 2560
• Includes shared metadata, embeddings, output tensors, per-layer GGUF artifacts, and the official multimodal projector
• Source quantization is preserved; this is a packaging transformation, not a new quantization

The package was validated against the original source GGUF for exact owned-tensor coverage, artifact sizes, layer coverage, and SHA-256 checksums. Skippy preflight also validated all package artifacts and a four-stage split plan.

AI-MAX startup

"$HOME/Library/Application Support/Flushnet/AI-MAX/bin/ai-max-local-model-webchat" \
  "LucianoM39/Gemma-4-E4B-it-Q4_0-Mesh-LLM"

Private/local Mesh-LLM split mode:

"$HOME/Library/Application Support/Flushnet/AI-MAX/bin/ai-max-local-model-webchat" \
  --runtime mesh-local \
  "LucianoM39/Gemma-4-E4B-it-Q4_0-Mesh-LLM"

AI-MAX uses a dedicated Gemma 4 adaptation profile with an 8192-token runtime context and a 2048-token response cap by default. These values can be overridden using the standard AI-MAX environment variables.

This repository contains the split package artifacts only; it does not duplicate the original monolithic source GGUF.