CogVLM2 19B needs ~21.8 GB VRAM. Mac Studio M1 Ultra 64GB has 46.1 GB. With Q4_K_M quantization, expect ~41 tok/s.
Operating mode
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.
Select quantization to explore
Fit status
Runs well
Decode
40.8 tok/s
TTFT
4744 ms
Safe context
8K
Memory
21.8 GB / 46.1 GB
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.
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | A | Runs well | 40.8 tok/s | 2588 ms | 8K |
| Coding | A | Runs well | 40.8 tok/s | 4744 ms | 8K |
| Agentic Coding | A | Runs well | 40.8 tok/s | 6900 ms | 8K |
| Reasoning | A | Runs well | 40.8 tok/s | 5606 ms | 8K |
| RAG | A | Runs well | 40.8 tok/s | 8625 ms | 8K |
How CogVLM2 19B (19B params) fits at each quantization level on Mac Studio M1 Ultra 64GB (46.1 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 7.4 GB | Low | A76 |
Q3_K_S | 3 | 9.3 GB | Low | A76 |
NVFP4 | 4 | 10.6 GB | Medium | A77 |
Q4_K_M | 4 | 11.6 GB | Medium | A77 |
Q5_K_M | 5 | 13.7 GB | High | A77 |
Q6_K | 6 | 15.6 GB | High | A78 |
Q8_0 | 8 | 20.3 GB | Very High | A80 |
F16Best for your GPU | 16 | 38.9 GB | Maximum | A81 |
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 99Your hardware
| Model | Params | Grade | Decode | Capabilities |
|---|---|---|---|---|
 Qwen3-Coder 30B A3B Instruct | 30.5B | S | 66.5 tok/s | |
| Qwen 3.5 27B | 27B | S | 28.9 tok/s | |
| Qwen 3.6 27B | 27B | S | 21.9 tok/s | |
| Qwen 3.6 35B A3B | 35B | S | 55.9 tok/s | |
| Qwen3-VL 30B A3B Instruct | 30B | S | 68.8 tok/s |
Yes, Mac Studio M1 Ultra 64GB can run CogVLM2 19B with a A grade (Runs well). Expected decode speed: 40.8 tok/s.
CogVLM2 19B (19B parameters) requires approximately 21.8 GB of memory with Q4_K_M quantization.
The recommended quantization for CogVLM2 19B is Q4_K_M, which balances quality and memory efficiency.
On Mac Studio M1 Ultra 64GB, CogVLM2 19B achieves approximately 40.8 tokens per second decode speed with a time-to-first-token of 4744ms using Q4_K_M quantization.
For coding workloads, CogVLM2 19B on Mac Studio M1 Ultra 64GB receives a A grade with 40.8 tok/s and 8K context.
On Mac Studio M1 Ultra 64GB, 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.
Not always. Mac Studio M1 Ultra 64GB 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.
Paste this snippet into any page to show a live fit card.
<iframe src="https://willitrunai.com/embed/cogvlm2-19b-on-m1-ultra-64gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>
Preview: