Models › granite-docling-258M
granite-docling-258M: VRAM needs and speed by GPU
No measured runs yet. Speeds below are estimates. Submit a run.
0.16B parameters · license: apache-2.0 · 8,192-token context · GGUF published Oct 5, 2025 · GGUF files on Hugging Face
Memory needed per quantization
| Quant | File size | Needs (estimate) | Smallest card it fits |
|---|---|---|---|
| Q8_0 | 0.2 GB | 1.0 GB | 8 GB cards |
| F16 | 0.3 GB | 1.1 GB | 8 GB cards |
Expected speed on popular hardware
| Hardware | Best quant that fits | Expected tok/s |
|---|---|---|
| NVIDIA RTX 3060 | Q8_0 | 257 estimate |
| NVIDIA RTX 4060 Ti | Q8_0 | 204 estimate |
| NVIDIA RTX 3090 | Q8_0 | 547 estimate |
| NVIDIA RTX 4090 | Q8_0 | 606 estimate |
| NVIDIA RTX 5090 | Q8_0 | 957 estimate |
| AMD Radeon RX 7900 XTX | Q8_0 | 608 estimate |
| Apple M4 Pro (20-core GPU) (64 GB) | Q8_0 | 193 estimate |
| Apple M3 Max (40-core GPU) (64 GB) | Q8_0 | 282 estimate |
| AMD Ryzen AI Max+ 395 (Radeon 8060S) | Q8_0 | 173 estimate |
Too big for your card?
Run granite-docling-258M on a rented 24-80 GB GPU by the hour.
Rent a bigger GPU on RunPodAffiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.