Gemma Tuner Multimodal vs Google Scion

Gemma Tuner Multimodal is an open-source fine-tuning toolkit for Google's Gemma 4 and Gemma 3n models that runs entirely on Apple Silicon using PyTorch with Metal Performance Shaders (MPS) backend — no NVIDIA GPU or cloud infrastructure required. It supports LoRA training on multimodal inputs: audio, images, and text simultaneously, using local CSV files or streamed from Google Cloud Storage or BigQuery. The tool targets the growing segment of developers who own M-series Macs but have been locked out of fine-tuning workflows that assume CUDA availability. Gemma 4's architecture is particularly well-suited to this use case: its 4B multimodal variant (designed for on-device deployment) trains efficiently on M3 Max and M4 Pro hardware within the available unified memory constraints. Primary use cases include medical transcription fine-tuning (audio → text with clinical terminology), visual QA systems (image + text → structured response), and private on-device pipelines where cloud API calls are prohibited by compliance requirements. The project fills a specific niche that Google's own fine-tuning documentation doesn't cover well for Apple hardware.

Google Scion is an open-source "hypervisor for agents" — a runtime that manages groups of AI agents in isolated containers, each with its own identity, credentials, git worktree, and toolset. Think of it as Kubernetes for agent teams: you declare your agent topology, Scion provisions the sandboxes, and agents can collaborate through structured channels without sharing file system or credential state. The isolation-over-constraints philosophy is Scion's core bet: rather than trying to constrain what a single powerful agent can do, give each agent a minimal, scoped environment where the blast radius of any failure or misbehavior is bounded. Harness adapters allow integration with Claude Code, Gemini CLI, and other existing agent runtimes — Scion acts as the orchestration layer above any underlying agent technology. For teams building multi-agent systems at scale, the credential isolation alone is a major feature — no more worrying about one agent leaking API keys to another. The Docker/Kubernetes support means it drops into existing infrastructure. Scion represents Google's opinionated answer to the question every AI platform team is grappling with: how do you run multiple AI agents safely in production without building a custom isolation layer from scratch?