Can CogVLM2 19B run on MacBook Air M3 24GB?

YES — With Offload

A79Great

Estimated from fit model

CogVLM2 19B needs ~17.5 GB VRAM. MacBook Air M3 24GB has 17.3 GB. With Q4_K_M quantization, expect ~6 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: Very lowStack: StandardBottleneck: Memory bandwidth

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) — 17.5 GB, 6.1 tok/s, Runs with offload (needs ~0.2 GB host RAM)

17.5 GB required17.3 GB available

101% VRAM needed

0.2 GB over capacity — needs offload or smaller quantization

Fit status

Runs with offload (needs ~0.2 GB host RAM)

Decode

6.1 tok/s

TTFT

31732 ms

Safe context

Memory

17.5 GB / 17.3 GB

Memory breakdown

Weights11.6 GB

KV Cache2.4 GB

Runtime0.9 GB

Headroom2.6 GB

See how fast it feels

See how fast it feelsCogVLM2 19B on MacBook Air M3 24GB
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: 6.1 tok/s decode · 31.7s TTFT (warm) · 15 tok/s prefill

What limits this setup

The model fits in shared memory, but shared-memory bandwidth is now the real limiter.

Fit does not mean dedicated-VRAM speed

Unified or shared memory can make a model technically fit, but sustained tokens per second may still trail a discrete high-bandwidth GPU with less total memory.

Very little memory headroom

You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.

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

Prioritize bandwidth, not only capacity

If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.

Buy headroom, not only minimum fit

A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	A	Tight fit	6.3 tok/s	16743 ms	8K
Coding	A	Runs with offload (needs ~0.2 GB host RAM)	6.1 tok/s	31732 ms	8K
Agentic Coding	B	Very compromised (needs ~1.6 GB host RAM)	5.0 tok/s	55870 ms	8K
Reasoning	A	Runs with offload (needs ~0.2 GB host RAM)	6.1 tok/s	37501 ms	8K
RAG	B	Very compromised (needs ~1.6 GB host RAM)	5.0 tok/s	69838 ms	8K

Quantization options

How CogVLM2 19B (19B params) fits at each quantization level on MacBook Air M3 24GB (17.3 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	7.4 GB	Low	A84
Q3_K_S	3	9.3 GB	Low	A84
NVFP4	4	10.6 GB	Medium	A84
Q4_K_MBest for your GPU	4	11.6 GB	Medium	A84
Q5_K_M	5	13.7 GB	High	F0
Q6_K	6	15.6 GB	High	F0
Q8_0	8	20.3 GB	Very High	F0
F16	16	38.9 GB	Maximum	F0

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 MacBook Air M3 24GB can run

Model	Params	Grade	Decode
Magistral Small 2507	24B	B	3.8 tok/s
Devstral Small 2 24B Instruct	24B	B	3.8 tok/s
Devstral Small 1.1	24B	B	3.8 tok/s
GPT-OSS 20B	21B	A	11.4 tok/s
Codestral 2 25.08	22B	B	4.3 tok/s

Frequently asked questions

Can MacBook Air M3 24GB run CogVLM2 19B?

Yes, MacBook Air M3 24GB can run CogVLM2 19B with a A grade (Runs with offload (needs ~0.2 GB host RAM)). Expected decode speed: 6.1 tok/s.

How much VRAM does CogVLM2 19B need?

CogVLM2 19B (19B parameters) requires approximately 17.5 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 MacBook Air M3 24GB?

On MacBook Air M3 24GB, CogVLM2 19B achieves approximately 6.1 tokens per second decode speed with a time-to-first-token of 31732ms using Q4_K_M quantization.

Can MacBook Air M3 24GB run CogVLM2 19B for coding?

For coding workloads, CogVLM2 19B on MacBook Air M3 24GB receives a A grade with 6.1 tok/s and 8K context.

What context window can CogVLM2 19B use on MacBook Air M3 24GB?

On MacBook Air M3 24GB, 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.

What should I upgrade first if CogVLM2 19B feels slow on MacBook Air M3 24GB?

Prioritize bandwidth, not only capacity. If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.

Is unified memory on MacBook Air M3 24GB as fast as VRAM for CogVLM2 19B?

Not always. MacBook Air M3 24GB 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 MacBook Air M3 24GB 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-air-24gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: