Can it run?
Runs well
Using Q6_K in Ollama
Fit status
Runs well
Decode
67.2 tok/s
TTFT
2881 ms
Safe context
24K
Memory
2.6 GB / 4.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | B | Runs well | 67.2 tok/s | 4190 ms | 48K |
| Chat | B | Runs well | 67.2 tok/s | 1571 ms | 12K |
| Coding | B | Runs well | 67.2 tok/s | 2881 ms | 24K |
| RAG | B | Runs well | 67.2 tok/s | 5238 ms | 48K |
| Reasoning | B | Runs well | 67.2 tok/s | 3405 ms | 24K |
How embeddinggemma 300M (0.30000001192092896B params) fits at each quantization level on GTX 1650 4GB (4.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 0.1 GB | Low | D30 |
Q3_K_S | 3 | 0.1 GB | Low | D30 |
NVFP4 | 4 | 0.2 GB | Medium | D30 |
Q4_K_M | 4 | 0.2 GB | Medium | D30 |
Q5_K_M | 5 | 0.2 GB | High | D30 |
Q6_K | 6 | 0.2 GB | High | D30 |
Q8_0 | 8 | 0.3 GB | Very High | D31 |
F16Best for your GPU | 16 | 0.6 GB | Maximum | D33 |
huggingface-cli download hf-ggml-org--embeddinggemma-300m-gguf