Models › Qwen3-8B-Base
Qwen3-8B-Base: VRAM needs and speed by GPU
No measured runs yet. Speeds below are estimates. Submit a run.
8.19B parameters · license: apache-2.0 · 32,768-token context · GGUF published Jul 28, 2026 · GGUF files on Hugging Face
Memory needed per quantization
| Quant | File size | Needs (estimate) | Smallest card it fits |
|---|---|---|---|
| Q8_0 | 8.7 GB | 9.7 GB | 12 GB cards |
| BF16 | 16.4 GB | 17.4 GB | 24 GB cards |
Expected speed on popular hardware
| Hardware | Best quant that fits | Expected tok/s |
|---|---|---|
| NVIDIA RTX 3060 | Q8_0 | 35.4 estimate |
| NVIDIA RTX 4060 Ti | Does not fit | |
| NVIDIA RTX 3090 | Q8_0 | 75.3 estimate |
| NVIDIA RTX 4090 | Q8_0 | 83.4 estimate |
| NVIDIA RTX 5090 | Q8_0 | 132 estimate |
| AMD Radeon RX 7900 XTX | Q8_0 | 83.8 estimate |
| Apple M4 Pro (20-core GPU) (64 GB) | Q8_0 | 26.6 estimate |
| Apple M3 Max (40-core GPU) (64 GB) | Q8_0 | 38.8 estimate |
| AMD Ryzen AI Max+ 395 (Radeon 8060S) | Q8_0 | 23.9 estimate |
Too big for your card?
Run Qwen3-8B-Base 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.