CogVLM2 19B needs ~16.9 GB VRAM. RX 7900 XT 20GB has 20.0 GB. With Q4_K_M quantization, expect ~45 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
Tight fit
Decode
44.5 tok/s
TTFT
4348 ms
Safe context
8K
Memory
16.9 GB / 20.0 GB
This setup is broadly balanced for this model.
No major red flags
This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | S | Runs well | 44.5 tok/s | 2372 ms | 8K |
| Coding | A | Tight fit | 44.5 tok/s | 4348 ms | 8K |
| Agentic Coding | A | Runs with offload | 44.5 tok/s | 6325 ms | 8K |
| Reasoning | A | Tight fit | 44.5 tok/s | 5139 ms | 8K |
| RAG | A | Runs with offload | 44.5 tok/s | 7906 ms | 8K |
Inference speed
Estimated decode speed (tokens/sec) for CogVLM2 19B at Q4_K_M across popular GPUs and Apple Silicon, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~105 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.
| GPU / Mac | Memory | Quant | Speed (tok/s) | Fits? |
|---|---|---|---|---|
 RTX 5090 32GB | 32 GB | Q4_K_M | 104.6 | Fits |
RX 7900 XTX 24GB | 24 GB | Q4_K_M | 64.1 | Fits |
| RTX 4090 24GB | 24 GB | Q4_K_M | 59.3 | Fits |
| RTX 3090 24GB | 24 GB | Q4_K_M | 54.5 | Fits |
Mac Studio M3 Ultra 256GB | 256 GB | Q4_K_M | 51.7 | Fits |
Mac Studio M2 Ultra 128GB | 128 GB | Q4_K_M | 43.0 | Fits |
Mac Studio M1 Ultra 128GB | 128 GB | Q4_K_M | 40.8 | Fits |
MacBook Pro M4 Max 128GB | 128 GB | Q4_K_M | 38.7 | Fits |
MacBook Pro M4 Max 64GB | 64 GB | Q4_K_M | 38.7 | Fits |
| RTX 4080 Super 16GB | 16 GB | Q4_K_M | 33.1 | Offloads |
MacBook Pro M4 Pro 48GB | 48 GB | Q4_K_M | 24.4 | Fits |
MacBook Pro M3 Max 64GB | 64 GB | Q4_K_M | 22.3 | Fits |
MacBook Pro M1 Max 64GB | 64 GB | Q4_K_M | 20.4 | Fits |
| RTX 4070 12GB | 12 GB | Q4_K_M | 11.8 | Too big |
| RTX 3060 12GB | 12 GB | Q4_K_M | 7.9 | Too big |
| RTX 4060 8GB | 8 GB | Q4_K_M | 2.8 | Too big |
Estimates for single-stream decoding at Q4_K_M; real tokens/sec varies with prompt length, context, batch size, and runtime build. Prompt processing (prefill) is faster than the decode figures shown here.
How CogVLM2 19B (19B params) fits at each quantization level on RX 7900 XT 20GB (20.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 7.4 GB | Low | A82 |
Q3_K_S | 3 | 9.3 GB | Low | A84 |
NVFP4 | 4 | 10.6 GB | Medium | A84 |
Q4_K_M | 4 | 11.6 GB | Medium | A84 |
Q5_K_M | 5 | 13.7 GB | High | A83 |
Q6_KBest for your GPU | 6 | 15.6 GB | High | A83 |
Q8_0 | 8 | 20.3 GB | Very High | F0 |
F16 | 16 | 38.9 GB | Maximum | F0 |
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 | A | 40.7 tok/s | |
| Qwen 3.5 27B | 27B | A | 18.3 tok/s | |
| Qwen 3.6 27B | 27B | S | 17.3 tok/s | |
| Qwen3-VL 30B A3B Instruct | 30B | A | 43.3 tok/s | |
 Magistral Small 2507 | 24B | S | 35.2 tok/s |
Yes, RX 7900 XT 20GB can run CogVLM2 19B with a A grade (Tight fit). Expected decode speed: 44.5 tok/s.
CogVLM2 19B (19B parameters) requires approximately 16.9 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 RX 7900 XT 20GB, CogVLM2 19B achieves approximately 44.5 tokens per second decode speed with a time-to-first-token of 4348ms using Q4_K_M quantization.
For coding workloads, CogVLM2 19B on RX 7900 XT 20GB receives a A grade with 44.5 tok/s and 8K context.
On RX 7900 XT 20GB, 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.
Paste this snippet into any page to show a live fit card.
<iframe src="https://willitrunai.com/embed/cogvlm2-19b-on-rx-7900-xt-20gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>
Preview: