gemma3-4b-qat-4bit
Run Gemma 3 4B (QAT) on a Mac
Run it
Install: pip install rapid-mlx or Rapid-MLX Desktop.
rapid-mlx serve gemma3-4b-qat-4bit
Minimum RAM: No published minimum; 7.6 GB measured peak. Leave macOS and context headroom. Agent setup guide
Google's quantisation-aware-trained Gemma 3 4B — the QAT checkpoint was trained to be quantised, so it loses less than a post-hoc 4-bit conversion. Takes image inputs too.
4 B dense · QAT 4-bit · visionvision + text on small Macs · Google-ecosystem prompts
Google 128K context Base model intelligence 1 AA
First request
rapid-mlx serve gemma3-4b-qat-4bit
Weights (3.03 GB) download on first run; you get an OpenAI-compatible
endpoint at http://localhost:8000/v1. No rapid-mlx yet? It's one line —
curl -fsSL https://rapidmlx.com/install.sh | bash — or take the
desktop app.
Point a client at it. OpenAI SDK: base_url="http://localhost:8000/v1",
model="gemma3-4b-qat-4bit". Claude Code:
ANTHROPIC_BASE_URL=http://localhost:8000 (no /v1 suffix) —
see the Claude Code guide.
Measured on real hardware
Headline machine · Mac mini M4 Pro · 48 GB
All measured machines
| Machine | Decode | First token | Peak memory | Cold boot | Weights |
|---|---|---|---|---|---|
| Mac mini M4 Pro · 48 GB rapid-mlx 0.15.5 · 2026-10-03 raw data | 91.3 tok/s | 2.7 s | 7.6 GB | — | 3.03 GB |
Each row cites its own run; numbers are never copied across machines. Mac mini M4 Pro · 48 GB: rapid-mlx 0.15.5, 2026-10-03 — raw data + method — 4 concurrent streams: 205.0 tok/s aggregate.
Will it fit your Mac?
Measured 7.6 GB Metal peak (4.6 GB process RSS) on a 48 GB M4 Pro across the whole benchmark run — up to a ~16K-token prompt and four concurrent streams, including the engine's prefix cache, which defaults to 20% of RAM. A short chat needs less; long contexts grow the KV cache, so leave headroom.
For the conservative install-default placement see the hardware tiers table; to compare against every model your RAM can hold, use the live picker.
Variants & profiles
gemma-4-12b-4bit— Gemma 4 12B — the current generation, dense
FAQ
How much memory does Gemma 3 4B (QAT) need on a Mac?
Measured 7.6 GB Metal peak (4.6 GB process RSS) on a 48 GB M4 Pro across the whole benchmark run — up to a ~16K-token prompt and four concurrent streams, including the engine's prefix cache, which defaults to 20% of RAM. A short chat needs less; long contexts grow the KV cache, so leave headroom.
How fast is Gemma 3 4B (QAT) on Apple Silicon?
We measured 91.3 tok/s on a Mac mini M4 Pro · 48 GB, rapid-mlx 0.15.5, 2026-10-03; peak memory 7.6 GB on that run. Every row on this page keeps its own chip, version and date — no number is copied across machines.
How do I run Gemma 3 4B (QAT) locally?
rapid-mlx serve gemma3-4b-qat-4bit — the weights download on first run and you get an OpenAI-compatible endpoint at localhost:8000/v1 for Aider, Codex and any OpenAI client, plus the Anthropic API at localhost:8000 for Claude Code.