Memory at each precision
| Version | Weights | With context | Smallest Mac |
|---|---|---|---|
| 8-bit | 116 GB | 125 GB | 192 GB |
| 6-bit | 88.6 GB | 95.7 GB | 128 GB |
| 4-bit | 61.3 GB | 66.2 GB | 96 GB |
| 3-bit lower quality | 47.7 GB | 51.5 GB | 96 GB |
| 2-bit lower quality | 34.1 GB | 36.8 GB | 64 GB |
On every Apple chip
The smallest memory size of each chip that runs Llama 4 Scout 17B-16E Instruct, the best version that fits it, and the writing speed you can expect.
| Chip | Needs | Version | Tokens/s |
|---|---|---|---|
| M6 | no size fits | Too big | – |
| M5 Ultra | 96 GB | 4-bit Runs well | 56–100 |
| M5 Max | 128 GB | 6-bit Runs well | 20–36 |
| M5 Pro | 64 GB | 2-bit Runs well | 26–46 |
| M5 | no size fits | Too big | – |
| M4 Max | 128 GB | 6-bit Runs well | 18–32 |
| M4 Pro | 64 GB | 2-bit Runs well | 23–41 |
| M4 | no size fits | Too big | – |
| M3 Ultra | 96 GB | 4-bit Runs well | 38–68 |
| M3 Max | 96 GB | 4-bit Runs well | 14–25 |
| M3 Pro | no size fits | Too big | – |
| M3 | no size fits | Too big | – |
| M2 Ultra | 128 GB | 6-bit Runs well | 26–46 |
| M2 Max | 96 GB | 4-bit Runs well | 19–33 |
| M2 Pro | no size fits | Too big | – |
| M2 | no size fits | Too big | – |
| M1 Ultra | 128 GB | 6-bit Runs well | 26–46 |
| M1 Max | 64 GB | 2-bit Runs well | 34–60 |
| M1 Pro | no size fits | Too big | – |
| M1 | no size fits | Too big | – |
How to run it
- The easy way: download MLX Harness, free and open source. Search for Llama 4 Scout 17B-16E 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-4-Scout-17B-16E-Instruct-4bitBuild: mlx-community/Llama-4-Scout-17B-16E-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 4 Scout 17B-16E Instruct
| Made by | Meta |
|---|---|
| Parameters | 109B total · 17B active (mixture of experts) |
| Context | 10M tokens |
| Licence | Llama 4 Community License |
| Released | 2025-04 |
| Input | Text and images |
| Intelligence | 6.5 on the Artificial Analysis Intelligence Index |
| Weights | meta-llama/Llama-4-Scout-17B-16E-Instruct |
Questions
- How much memory does Llama 4 Scout 17B-16E Instruct need?
- About 61.3 GB of weights at 4-bit, plus room for the context: 66.2 GB in all. On a Mac that means 96 GB of unified memory or more, since macOS lets the GPU use about 75% of it by default.
- How fast is Llama 4 Scout 17B-16E 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 56–100 tokens a second (4-bit). The smallest Mac that runs it, with 96 GB, manages about 56–100 on an M5 Ultra.
- Why does Llama 4 Scout 17B-16E Instruct run faster than its size suggests?
- It is a mixture-of-experts model: it holds 109B parameters in memory but reads only 17B for each token, so it writes at the speed of a much smaller model.
Similar models
For your model card or README
Shows the smallest Mac that runs Llama 4 Scout 17B-16E Instruct well and links here.
[](https://mlx.app/run/llama-4-scout)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.