Skip to content

Phi 4 on M3 Ultra 96 GB: VRAM, speed and fit

Phi 4 at 8-bit runs on the M3 Ultra 96 GB with 16K of context at about 32 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on M3 Ultra 96 GB · 8-bit (Q8_0) · Ollama / LM Studio · Chat

GPU memory
42.2 GB
of 96.0 GB
Context
16K
auto
Speed
32 tok/s
26 when full
First reply
2.7 s
2K prompt

Where the memory goes

Model
14.5 GB
Context
3.1 GB
Overhead
612 MB
OS
24.0 GB
Free
53.8 GB
Memory
42.2 GB of 96.0 GB
Context
16K tokens · auto
Speed
32 tok/s · 26 when full
First reply
2.7 s for a 2K prompt
Running cost
$3.55/mo at home · $30/mo on a cloud RTX A4500

Validation: Validated

What Phi 4 on M3 Ultra 96 GB was checked against
CheckTargetResultStatus
Fits in GPU memory72.0 GB usable18.2 GB neededPass
Headroom5% free53.8 GB free (75%)Pass
Context for chat16K or more16KPass
Writing speed15 tok/s or more32 tok/sPass
First replyunder 3 s, 2K prompt2.7 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
Phi 4
Parameters
14.7B
Precision
8-bit (Q8_0)
Context
16K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
M3 Ultra 96 GB
GPU can use
72.0 GB of 96.0 GB

Running cost

At home
$3.55/mo
162 W at the wall
Renting instead
$30/mo
RTX A4500 at $0.25/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 21, 2026.

How to run it

llama.cpp
llama-server -hf MaziyarPanahi/phi-4-GGUF
vLLM
vllm serve microsoft/phi-4
MLX (Mac)
mlx_lm.chat --model mlx-community/phi-4-4bit

Phi 4 on similar machines

All machines: Phi 4 VRAM requirements. Everything that runs here: what can the M3 Ultra 96 GB run?

Other models that run on M3 Ultra 96 GB

Questions

Can Phi 4 run on M3 Ultra 96 GB?

Phi 4 at 8-bit runs on the M3 Ultra 96 GB with 16K of context at about 32 tokens a second.

Report VR-1FIHE74, calculated Nov 22, 2026 from published specs, not benchmarked.