Skip to content

Qwen2.5 14B Instruct 1M on M3 Ultra 96 GB: VRAM, speed and fit

Qwen2.5 14B Instruct 1M at 8-bit runs on the M3 Ultra 96 GB with 256K of context at about 31 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
87.3 GB
of 96.0 GB
Context
256K
auto
Speed
31 tok/s
7.8 when full
First reply
2.6 s
2K prompt

Where the memory goes

Model
14.6 GB
Context
48.0 GB
Overhead
713 MB
OS
24.0 GB
Free
8.7 GB
Memory
87.3 GB of 96.0 GB
Context
256K tokens · auto
Speed
31 tok/s · 7.8 when full
First reply
2.6 s for a 2K prompt
Running cost
$3.55/mo at home · $122/mo on a cloud 4× RTX A4500

Validation: Validated

What Qwen2.5 14B Instruct 1M on M3 Ultra 96 GB was checked against
CheckTargetResultStatus
Fits in GPU memory72.0 GB usable63.3 GB neededPass
Headroom5% free8.7 GB free (12%)Pass
Context for chat16K or more256KPass
Writing speed15 tok/s or more31 tok/sPass
First replyunder 3 s, 2K prompt2.6 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
Qwen2.5 14B Instruct 1M
Parameters
14.8B
Precision
8-bit (Q8_0)
Context
256K, 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
$122/mo
4× RTX A4500 at $1.00/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 lmstudio-community/Qwen2.5-14B-Instruct-1M-GGUF
vLLM
vllm serve Qwen/Qwen2.5-14B-Instruct-1M
MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen2.5-14B-Instruct-1M-8bit

Qwen2.5 14B Instruct 1M on similar machines

All machines: Qwen2.5 14B Instruct 1M 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 Qwen2.5 14B Instruct 1M run on M3 Ultra 96 GB?

Qwen2.5 14B Instruct 1M at 8-bit runs on the M3 Ultra 96 GB with 256K of context at about 31 tokens a second.

Report VR-09F409Y, calculated Nov 22, 2026 from published specs, not benchmarked.