Quadro M2000M holds 105 of the models in our catalogue and is, in practice, a Q4_K_M card — the largest it takes is InternLM2 5B at Q4_K_M.
Neighbours in memory rather than in price: memory decides whether a card can do the job at all, so two cards of the same size at different prices is the comparison you are making. Ordered by memory, then bandwidth — not by the value column, which is worked out from bandwidth and price alone and therefore rewards a cheap card whatever its software stack does to that bandwidth in practice. Read it as one input, not as a ranking.
NVIDIA Quadro M2000M — 4 GB VRAM.
- Brand
- NVIDIA
- VRAM
- 4 GB GDDR5
- Bandwidth
- 80 GB/s
- FP16 Compute
- 1.5 TFLOPS
- FP32 Compute
- 1.5 TFLOPS
- CUDA Cores
- 640
- TDP
- 55 W
- Architecture
- Maxwell
With 4 GB VRAM and 80 GB/s bandwidth, this GPU handles models up to 4.25B parameters.
Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~9 tok/s.
| Model | Size | VRAM Q4 | tok/s | Avg |
|---|---|---|---|---|
| Ministral 3 3B Reasoning | 4.25B | 3.1 GB | 15 | 60.1 |
| Qwen 1.5 4B | 4B | 2.9 GB | 16 | 12.6 |
| Qwen3 4B | 4B | 2.9 GB | 16 | 40.7 |
| Qwen3-4B Instruct 2507 | 4B | 2.9 GB | 16 | 37.2 |
| Qwen3-Embedding 4B | 4B | 2.9 GB | 16 | 68.5 |
| Nemotron 3 Nano 4B | 3.97B | 2.9 GB | 16 | 60.1 |
| Ministral 3 3B | 3.85B | 2.8 GB | 17 | 18.3 |
| Phi-3.5 Mini 3.8B | 3.82B | 2.8 GB | 17 | 46.6 |
| phi-3-mini-4k 3.8B | 3.8B | 2.8 GB | 17 | 30.5 |
| Phi-4-mini 3.8B | 3.8B | 2.8 GB | 17 | 49.0 |
| Granite 4.2 3B | 3.66B | 2.7 GB | 17 | 38.0 |
| Cogito 3B | 3.61B | 2.7 GB | 18 | 22.1 |
| Falcon3-3B | 3.23B | 2.5 GB | 20 | 25.7 |
| granite-4.0-h-micro 3.2B | 3.2B | 2.4 GB | 20 | 27.3 |
| Llama-3.2-3B | 3.2B | 2.4 GB | 20 | 17.9 |
| Falcon-H1 3B | 3.15B | 2.4 GB | 20 | 49.5 |
| LFM2.5-VL 3B | 3.12B | 2.4 GB | 21 | 53.3 |
| Qwen 2.5 3B | 3.1B | 2.4 GB | 21 | 37.2 |
| SmolLM3-3B | 3.1B | 2.4 GB | 21 | 35.1 |
| Ministral 3B | 3B | 2.3 GB | 21 | 23.7 |