Can internlm2 limarp chat 20b run on Mac Studio M3 Ultra 96GB?

YES — Runs Great

C47Usable

○Estimated from fit model

internlm2 limarp chat 20b needs ~25.8 GB VRAM. Mac Studio M3 Ultra 96GB has 69.1 GB. With Q4_K_M quantization, expect ~46 tok/s.

Runtime: llama.cppCapacity: RoomyBandwidth: HighStack: StandardBottleneck: Balanced

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 25.8 GB, 45.6 tok/s, Runs well

25.8 GB required69.1 GB available

37% VRAM used

Fit status

Runs well

Decode

45.6 tok/s

TTFT

4241 ms

Safe context

312K

Memory

25.8 GB / 69.1 GB

Memory breakdown

Weights12.2 GB

KV Cache2.3 GB

Runtime0.9 GB

Headroom10.4 GB

See how fast it feels

See how fast it feelsinternlm2 limarp chat 20b on Mac Studio M3 Ultra 96GB
1st promptCold start — includes initialization
>What is local AI inference?
Local AI inference means running an AI language model directly on your own hardware — your laptop, desktop, or server — instead of sending requests to a remote cloud API.

When you run inference locally the model weights are loaded into your GPU or unified memory. Each token you generate requires reading those weights from memory, so memory bandwidth is the main bottleneck for decode speed.

Key benefits of running locally:
 - Full privacy: your prompts never leave your machine
 - No per-token cost or rate limits
 - Works offline once the model is downloaded
 - Latency depends only on your hardware
2nd promptWarm — model ready, no init overhead
>How much VRAM do I need?
It depends on the model size and quantization level. A rough rule of thumb:

 Model size Q4 (4-bit) Q8 (8-bit) FP16
 7B params ~4.3 GB ~7.5 GB ~14 GB
 13B params ~7.9 GB ~13.9 GB ~26 GB
 70B params ~42.7 GB ~74.9 GB ~140 GB

Most people use 4-bit quantization (Q4_K_M) which gives 90-95% of full quality at a fraction of the memory. A 24 GB GPU can comfortably run most 7B-13B models.
Estimated: 45.6 tok/s decode · 4.2s TTFT (warm) · 114 tok/s prefill

What limits this setup

This setup is broadly balanced for this model.

Shared-memory contention still exists

The OS, browser, and inference runtime all compete for the same physical memory pool, so real-world headroom is less forgiving than raw capacity suggests.

Best improvement path

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	C	Runs well	45.6 tok/s	2313 ms	312K
Coding	C	Runs well	45.6 tok/s	4241 ms	312K
Agentic Coding	C	Runs well	45.6 tok/s	6169 ms	312K
Reasoning	C	Runs well	45.6 tok/s	5012 ms	312K
RAG	C	Runs well	45.6 tok/s	7711 ms	312K

Quantization options

How internlm2 limarp chat 20b (20B params) fits at each quantization level on Mac Studio M3 Ultra 96GB (69.1 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	7.8 GB	Low	C40
Q3_K_S	3	9.8 GB	Low	C40
NVFP4	4

Get started

Copy-paste commands to run internlm2 limarp chat 20b on your machine.

Run

lms load hf-intervitens-archive--internlm2-limarp-chat-20b-gguf && lms server start

Upgrade options

Hardware that runs internlm2 limarp chat 20b well

👁 NVIDIA
NVIDIA A100 80GBBudget pick

2039 GB/s (+1220)

Raises estimated decode speed by about 208%.140.4 tok/s decode

Raises estimated decode speed by about 208%.

~$15,000 MSRP

👁 NVIDIA
NVIDIA A800 80GBBest value

1935 GB/s (+1116)

Raises estimated decode speed by about 171%.123.7 tok/s decode

Raises estimated decode speed by about 171%.