Gemma 2

What is Gemma 2?

Gemma 2 is Google's open-weights language model. Released in 2024-07, it ships in 3 sizes (2B, 9B, 27B) and is licensed Gemma License. The most popular use case is Google-quality chat at small sizes.

VRAM and hardware

The smallest 2B size needs at least 2 GB VRAM; the largest 27B needs around 2 GB.

Size	Min VRAM (Q4_K_M)	Recommended VRAM	Tokens/sec on 3090
2B	2 GB	3 GB	~31 tok/s
9B	6 GB	9 GB	~24 tok/s
27B	17 GB	26 GB	~18 tok/s

How to run Gemma 2 locally

Option 1: Ollama (simplest)

ollama pull gemma-2
ollama run gemma-2

Option 2: Mullama (production)

mullama pull gemma-2
mullama run gemma-2

Option 3: llama.cpp (CLI)

# Download a GGUF from Hugging Face (search "gemma-2 gguf")
./llama-cli -m gemma-2.Q4_K_M.gguf -p "Hello, AI!"

Option 4: Python with Mullama or llama-cpp-python

from mullama import Model, Context
model = Model.load("gemma-2.Q4_K_M.gguf", n_gpu_layers=99)
ctx = Context(model, n_ctx=4096)
print(ctx.generate("Hello, AI!", 256))

What you can build with Gemma 2

• chat
• edge
• low-resource

Hardware it fits on

• Apple Silicon (8GB): 2B in MLX
• Apple Silicon (16GB): 9B in MLX
• Apple Silicon (32GB): 27B in MLX or GGUF
• Apple Silicon (64GB+): 27B in MLX
• RTX 3090 (24GB): 27B at Q4_K_M
• RTX 4090 (24GB): same as 3090, ~30% faster
• RTX 5090 (32GB): 27B at Q8_0
• 2× RTX 6000 Ada (96GB total): 27B at Q4_K_M

Related models in the Gemma family

• Gemma 3 — 270M, 1B, 4B, 12B, 27B · single-GPU multimodal chat
• Gemma 4 — e2b, e4b, 12b, 26b, 31b · frontier-class multimodal with strong reasoning

Other models you might consider

• Llama 3.1 (Meta) — 8B, general-purpose chat, agentic workflows, code
• Llama 3.2 (Meta) — 1B, low-end hardware, mobile, edge deployment
• Llama 3.2 Vision (Meta) — 11B, vision + chat on desktop GPUs
• Llama 3.3 (Meta) — 70B, best-in-class 70B for desktops
• Llama 4 (Meta) — Scout 17B (109B MoE), vision + long context (10M tokens)
• Qwen 2.5 (Alibaba) — 0.5B, strongest 7B-32B on consumer hardware