Memory at each precision
| Version | Weights | With context | Smallest Mac |
|---|---|---|---|
| 8-bit | 3.4 GB | 4.9 GB | 8 GB |
| 6-bit | 2.6 GB | 4.1 GB | 8 GB |
| 4-bit | 1.8 GB | 3.3 GB | 8 GB |
| 3-bit lower quality | 1.4 GB | 2.9 GB | 8 GB |
| 2-bit lower quality | 1 GB | 2.5 GB | 8 GB |
On every Apple chip
The smallest memory size of each chip that runs Llama 3.2 3B Instruct, the best version that fits it, and the writing speed you can expect.
| Chip | Needs | Version | Tokens/s |
|---|---|---|---|
| M6 | 16 GB | 8-bit Runs well | 20–36 |
| M5 Ultra | 96 GB | 8-bit Runs well | 159–282 |
| M5 Max | 36 GB | 8-bit Runs well | 61–108 |
| M5 Pro | 24 GB | 8-bit Runs well | 41–72 |
| M5 | 16 GB | 8-bit Runs well | 20–36 |
| M4 Max | 36 GB | 8-bit Runs well | 54–96 |
| M4 Pro | 24 GB | 8-bit Runs well | 36–64 |
| M4 | 16 GB | 8-bit Runs well | 16–28 |
| M3 Ultra | 96 GB | 8-bit Runs well | 108–193 |
| M3 Max | 36 GB | 8-bit Runs well | 40–71 |
| M3 Pro | 18 GB | 8-bit Runs well | 20–35 |
| M3 | 8 GB | 8-bit Runs well | 13–24 |
| M2 Ultra | 64 GB | 8-bit Runs well | 106–188 |
| M2 Max | 32 GB | 8-bit Runs well | 53–94 |
| M2 Pro | 16 GB | 8-bit Runs well | 26–47 |
| M2 | 8 GB | 8-bit Runs well | 13–24 |
| M1 Ultra | 64 GB | 8-bit Runs well | 106–188 |
| M1 Max | 32 GB | 8-bit Runs well | 53–94 |
| M1 Pro | 16 GB | 8-bit Runs well | 26–47 |
| M1 | 8 GB | 6-bit Runs well | 12–21 |
How to run it
- The easy way: download MLX Harness, free and open source. Search for Llama 3.2 3B Instruct, download the version that fits your Mac, and chat. It also serves the model to your other apps over an OpenAI-compatible API.Download MLX HarnessFree · Apple Silicon · macOS 14.5 or later
- In Terminal with Apple’s MLX:
pip install mlx-lmmlx_lm.chat --model mlx-community/Llama-3.2-3B-Instruct-4bitBuild: mlx-community/Llama-3.2-3B-Instruct-4bit on Hugging Face. Other precisions sit next to it.
Step by step: download and run LLMs on a Mac · how to use MLX Harness · MLX vs Ollama vs llama.cpp
About Llama 3.2 3B Instruct
| Made by | Meta |
|---|---|
| Parameters | 3.21B (dense) |
| Context | 128K tokens |
| Licence | Llama 3.2 Community License |
| Released | 2024-09 |
| Input | Text |
| Weights | meta-llama/Llama-3.2-3B-Instruct |
Questions
- How much memory does Llama 3.2 3B Instruct need?
- About 1.8 GB of weights at 4-bit, plus room for the context: 3.3 GB in all. On a Mac that means 8 GB of unified memory or more, since macOS lets the GPU use about 75% of it by default.
- How fast is Llama 3.2 3B Instruct on a Mac?
- Writing speed depends on memory bandwidth. The fastest Mac for it here is an M5 Ultra with 96 GB at about 159–282 tokens a second (8-bit). The smallest Mac that runs it, with 8 GB, manages about 13–24 on an M3.
- Does quantization hurt Llama 3.2 3B Instruct?
- At 4 bits most models stay close to the original quality; below 4 bits quality drops faster. Pick the highest precision that fits your Mac with room to spare.
Similar models
For your model card or README
Shows the smallest Mac that runs Llama 3.2 3B Instruct well and links here.
[](https://mlx.app/run/llama-3.2-3b)How it’s worked out
Memory: the model’s parameters at 8, 6, 4, 3 or 2 bits per weight plus MLX’s group scales (half a bit per weight), or its native format, with room for the context on top. By default macOS lets the GPU use about 75% of unified memory. Speed: memory bandwidth divided by the bytes read per token is the ceiling for one conversation; in llama.cpp’s public Apple Silicon benchmarks real runs reach about 45% to 80% of it, which is the range shown. These are estimates, not measurements: measure your own Mac on MLX Bench. Model figures from each official model card, chip figures from Apple’s Tech Specs. Checked 2026-10-19. Same data as mlx.app/compare.