Models › Mistral-7B-Instruct-v0.3
Mistral-7B-Instruct-v0.3: VRAM needs and speed by GPU
At Q4_K_M, Mistral-7B-Instruct-v0.3 needs about 5.4 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.
7.25B parameters · license: apache-2.0 · 32,768-token context · GGUF published May 22, 2024 · GGUF files on Hugging Face
Memory needed per quantization
| Quant | File size | Needs (estimate) | Smallest card it fits |
|---|---|---|---|
| IQ3_M | 3.3 GB | 4.3 GB | 8 GB cards |
| Q3_K_L | 3.8 GB | 4.8 GB | 8 GB cards |
| IQ4_NL | 4.1 GB | 5.1 GB | 8 GB cards |
| Q4_K_M | 4.4 GB | 5.4 GB | 8 GB cards |
| Q5_K_M | 5.1 GB | 6.1 GB | 8 GB cards |
| Q6_K | 5.9 GB | 6.9 GB | 8 GB cards |
| Q8_0 | 7.7 GB | 8.7 GB | 12 GB cards |
Expected speed on popular hardware
| Hardware | Best quant that fits | Expected tok/s |
|---|---|---|
| NVIDIA RTX 3060 | Q8_0 | 40.0 estimate |
| NVIDIA RTX 4060 Ti | Q6_K | 41.2 estimate |
| NVIDIA RTX 3090 | Q8_0 | 85.2 estimate |
| NVIDIA RTX 4090 | Q8_0 | 94.3 estimate |
| NVIDIA RTX 5090 | Q8_0 | 149 estimate |
| AMD Radeon RX 7900 XTX | Q8_0 | 94.7 estimate |
| Apple M4 Pro (20-core GPU) (64 GB) | Q8_0 | 30.1 estimate |
| Apple M3 Max (40-core GPU) (64 GB) | Q8_0 | 43.9 estimate |
| AMD Ryzen AI Max+ 395 (Radeon 8060S) | Q8_0 | 27.0 estimate |
Too big for your card?
Run Mistral-7B-Instruct-v0.3 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.