Can CogVLM2 19B run on Mac Studio M3 Ultra 96GB?

YES — Runs Great

A81Great

Estimated from fit model

CogVLM2 19B needs ~25.3 GB VRAM. Mac Studio M3 Ultra 96GB has 69.1 GB. With Q4_K_M quantization, expect ~48 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) — 25.3 GB, 51.7 tok/s, Runs well

25.3 GB required69.1 GB available

37% VRAM used

Fit status

Runs well

Decode

51.7 tok/s

TTFT

3748 ms

Safe context

Memory

25.3 GB / 69.1 GB

Memory breakdown

Weights11.6 GB

KV Cache2.4 GB

Runtime0.9 GB

Headroom10.4 GB

See how fast it feels

See how fast it feelsCogVLM2 19B 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: 51.7 tok/s decode · 3.7s TTFT (warm) · 129 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	A	Runs well	48.1 tok/s	2198 ms	8K
Coding	A	Runs well	48.1 tok/s	4029 ms	8K
Agentic Coding	A	Runs well	48.1 tok/s	5860 ms	8K
Reasoning	A	Runs well	48.1 tok/s	4762 ms	8K
RAG	A	Runs well	48.1 tok/s	7325 ms	8K

Quantization options

How CogVLM2 19B (19B params) fits at each quantization level on Mac Studio M3 Ultra 96GB (69.1 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	7.4 GB	Low	A74
Q3_K_S	3	9.3 GB	Low	A74
NVFP4	4	10.6 GB	Medium	A74
Q4_K_M	4	11.6 GB	Medium	A74
Q5_K_M	5	13.7 GB	High	A75
Q6_K	6	15.6 GB	High	A75
Q8_0	8	20.3 GB	Very High	A76
F16Best for your GPU	16	38.9 GB	Maximum	A81

Get started

Copy-paste commands to run CogVLM2 19B on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "THUDM/cogvlm2-llama3-chat-19B" \
  --hf-file "cogvlm2-llama3-chat-19B-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Your hardware

More models your Mac Studio M3 Ultra 96GB can run

Model	Params	Grade	Decode
Qwen3-Coder 30B A3B Instruct	30.5B	S	84.2 tok/s
Qwen 3.5 27B	27B	S	36.5 tok/s
Qwen 3.6 27B	27B	S	27.8 tok/s
Qwen 3.6 35B A3B	35B	S	70.8 tok/s
Qwen3-VL 30B A3B Instruct	30B	S	87.1 tok/s

Frequently asked questions

Can Mac Studio M3 Ultra 96GB run CogVLM2 19B?

Yes, Mac Studio M3 Ultra 96GB can run CogVLM2 19B with a A grade (Runs well). Expected decode speed: 48.1 tok/s.

How much VRAM does CogVLM2 19B need?

CogVLM2 19B (19B parameters) requires approximately 25.3 GB of memory with Q4_K_M quantization.

What is the best quantization for CogVLM2 19B?

The recommended quantization for CogVLM2 19B is Q4_K_M, which balances quality and memory efficiency.

What speed will CogVLM2 19B run at on Mac Studio M3 Ultra 96GB?

On Mac Studio M3 Ultra 96GB, CogVLM2 19B achieves approximately 48.1 tokens per second decode speed with a time-to-first-token of 4029ms using Q4_K_M quantization.

Can Mac Studio M3 Ultra 96GB run CogVLM2 19B for coding?

For coding workloads, CogVLM2 19B on Mac Studio M3 Ultra 96GB receives a A grade with 48.1 tok/s and 8K context.

What context window can CogVLM2 19B use on Mac Studio M3 Ultra 96GB?

On Mac Studio M3 Ultra 96GB, CogVLM2 19B can safely use up to 8K tokens of context. The model's official context limit is 8K, but available memory constrains the safe maximum.

Is unified memory on Mac Studio M3 Ultra 96GB as fast as VRAM for CogVLM2 19B?

Not always. Mac Studio M3 Ultra 96GB can often fit larger models thanks to unified memory, but a discrete GPU with dedicated high-bandwidth VRAM may still decode faster once the model fits. For this combination, the important distinction is capacity versus sustained throughput.

See all results for Mac Studio M3 Ultra 96GB See all hardware for CogVLM2 19B

Embed this result▼

Paste this snippet into any page to show a live fit card.

<iframe src="https://willitrunai.com/embed/cogvlm2-19b-on-m3-ultra-96gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: