Run Gemma 4 31B on Apple Silicon
Gemma 4 31B is Google's largest dense Gemma 4 model, released on 2 April 2026 under Apache 2.0, with a 256K context and image input. At Q4_K_M the weights are about 19 GB, so a Mac with 32 GB runs it and 48 GB or more leaves room for a long context.
How much memory does Gemma 4 31B need?
Weight size is parameter count × bits per weight; recommended memory adds 15% runtime headroom plus a KV-cache allowance for the 262,144-token context window.
| Quantization | Weights | Recommended RAM | Cheapest fit | Notes |
|---|---|---|---|---|
| Q4_K_M | ~19 GB | 32 GB | M4 (Mac mini) | Best quality/size trade-off — the default for most deployments |
| Q8_0 | ~33.3 GB | 48 GB | M4 Pro (Mac mini) | Near-lossless; noticeably better on code and long reasoning |
| FP16 | ~62.6 GB | 80 GB | M4 Max (Mac Studio) | Full precision — research and eval baselines |
Which Mac should you pick?
Generation speed on Apple Silicon scales mostly with memory bandwidth, so the chip choice sets both whether the model fits and how fast it runs.
| Chip | Max RAM | Bandwidth | Fits Gemma 4 31B? | From |
|---|---|---|---|---|
| M4 · Mac mini | 32 GB | 120 GB/s | Q4 | $99/mo |
| M4 Pro · Mac mini | 64 GB | 273 GB/s | Q4 + Q8 | $200/mo |
| M4 Max · Mac Studio | 128 GB | 546 GB/s | Q4 + Q8 + FP16 | $286/mo |
| M3 Ultra · Mac Studio | 256 GB | 819 GB/s | Q4 + Q8 + FP16 | $571/mo |
Deploy Gemma 4 31B in about 5 minutes
- 1
Build a Mac in the constructor — pick an M4 with at least 32 GB of unified memory.
- 2
Pick any Ollama-based stack (or a clean machine) and pull Gemma 4 31B with one command — every deployment ships the OpenAI-compatible API.
- 3
Point your existing OpenAI SDK code at the new base_url and ship.
from openai import OpenAI
client = OpenAI(
api_key="mcy_live_...",
base_url="https://dep-xxxx.macyou.cloud/v1",
)
response = client.chat.completions.create(
model="gemma-4-31b",
messages=[{"role": "user", "content": "Hello!"}],
)License and what it's good at
Apache 2.0 — Apache 2.0: commercial use is allowed under the standard license terms. Made by Google.
- Apache 2.0 license, a first for Gemma
- Text and image input with a 256K context
- Fits a 32 GB Mac at Q4
Typical workloads: a private general assistant on one mac; multilingual writing and translation; reading documents and screenshots.
Frequently asked questions
Gemma 4 31B or Gemma 4 26B A4B?
The 31B is dense and reads every weight for every token. The 26B A4B is a mixture of experts with about 4B active, so it generates much faster on the same Mac.