Llama Meta Released 2024-09 chatsmallmobile

Llama 3.2

Meta's smallest Llama 3 models. 1B fits in 1GB VRAM; 3B is a sweet spot for low-end desktops and laptops.

Best for low-end hardware, mobile, edge deployment
Sizes 1B · 3B
Context 128K
License Llama 3.2 Community License
Min VRAM (default size, Q4) 1 GB
Rec VRAM 4 GB

What is Llama 3.2?

Llama 3.2 is Meta's open-weights language model. Released in 2024-09, it ships in 2 sizes (1B, 3B) and is licensed Llama 3.2 Community License. The most popular use case is low-end hardware, mobile, edge deployment.

VRAM and hardware

The smallest 1B size needs at least 1 GB VRAM; the largest 3B needs around 1 GB.

Size Min VRAM (Q4_K_M) Recommended VRAM Tokens/sec on 3090
1B 1 GB 2 GB ~34 tok/s
3B 2 GB 3 GB ~31 tok/s

How to run Llama 3.2 locally

Option 1: Ollama (simplest)

ollama pull llama-3-2
ollama run llama-3-2

Option 2: Mullama (production)

mullama pull llama-3-2
mullama run llama-3-2

Option 3: llama.cpp (CLI)

# Download a GGUF from Hugging Face (search "llama-3-2 gguf")
./llama-cli -m llama-3-2.Q4_K_M.gguf -p "Hello, AI!"

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

from mullama import Model, Context
model = Model.load("llama-3-2.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 Llama 3.2

  • chat
  • edge
  • mobile
  • low-resource

Hardware it fits on

  • Apple Silicon (8GB): 1B in MLX
  • Apple Silicon (16GB): 3B in MLX
  • Apple Silicon (32GB): 3B in MLX or GGUF
  • Apple Silicon (64GB+): 3B in MLX
  • RTX 3090 (24GB): 3B at Q4_K_M
  • RTX 4090 (24GB): same as 3090, ~30% faster
  • RTX 5090 (32GB): 3B at Q8_0
  • 2× RTX 6000 Ada (96GB total): 3B at Q4_K_M

Related models in the Llama family

  • Llama 3.1 — 8B, 70B, 405B · general-purpose chat, agentic workflows, code
  • Llama 3.2 Vision — 11B, 90B · vision + chat on desktop GPUs
  • Llama 3.3 — 70B · best-in-class 70B for desktops
  • Llama 4 — Scout 17B (109B MoE), Maverick 17B (400B MoE) · vision + long context (10M tokens)

Other models you might consider

  • Qwen 2.5 (Alibaba) — 0.5B, strongest 7B-32B on consumer hardware
  • Qwen 2.5-Coder (Alibaba) — 0.5B, best open-weight code model in 2024-2025
  • Qwen 3 (Alibaba) — 0.6B, best all-round local LLM in 2025-2026
  • Qwen 3-Coder (Alibaba) — 30B, long-context code agent workflows
  • DeepSeek R1 (DeepSeek) — 1.5B, reasoning, math, step-by-step problem solving
  • DeepSeek V3 (DeepSeek) — 671B (37B active MoE), frontier-class open-weights LLM (needs data-center GPUs)