Run Llama 4 Scout on M4 Ultra

Yes — Llama 4 Scout (109B) runs at 30 tok/s on M4 Ultra with 192 GB RAM using Q4_K_M quantization via MLX. First token latency is 1.3s. Meta's 109B Llama 4 Scout MoE model with a 10M-token context window.

Speed

tok/s

First Token

1.3

seconds

RAM Needed

192

GB minimum

Engine

MLX

recommended

Benchmark Details

LLMCheck measured Llama 4 Scout on M4 Ultra using the standard methodology: Q4_K_M quantization, 256-token input, 512-token output, 3 runs averaged on a freshly-booted system.

Metric	Value
Tokens per second	30 tok/s
Time to first token	1.3s
Quantization	Q4_K_M
Minimum RAM	192 GB
Recommended engine	MLX
Parameters	109B
Benchmark date	2026-02

Q4_K_M 109B MLX M4 Ultra

Setup Guide: Run Llama 4 Scout on M4 Ultra

The recommended engine for Llama 4 Scout on M4 Ultra is MLX. Install with pip and pull the model:

pip install mlx-lm

mlx_lm.generate --model mlx-community/llama-4-scout-q4_k_m --prompt "Hello!"

Alternatively, you can use Ollama for a simpler setup:

ollama run llama4:scout

Performance on Other Apple Silicon Chips

Chip	Speed	First Token	Min RAM	Engine
M5 Max	26 tok/s	1.5s	128 GB	MLX
M5 Max	22 tok/s	1.8s	128 GB	Ollama

System Requirements

To run Llama 4 Scout on M4 Ultra you need:

• Mac with M4 Ultra chip (or newer)
• 192 GB unified memory minimum
• macOS 13 Ventura or later
• ~163-192 GB free disk space for the model file
• MLX installed (see our Ollama install guide)

Compare More Models

See how Llama 4 Scout stacks up against other models on your specific Mac hardware.

Open Compare Tool Full Leaderboard