Moistral-11B-v3
by TheDrummer
10.7B params · text-generation · 110 likes · 113 downloads
Moistral-11B-v3 is a 10.7B parameter model. At Q4 quantization it requires 5GB of VRAM. It runs comfortably on GeForce RTX 4090 (122 tok/s), GeForce RTX 5090 (183 tok/s), M4 Max 128GB (44 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| Featherless |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 122 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 183 tok/s | comfortable |
| M4 Max 128GB | 128GB | 44 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 22 tok/s | tight |
| M4 Pro 24GB | 24GB | 22 tok/s | tight |
| A100 PCIe 80 GB | 80GB | 224 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 450 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 108 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 89 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 86 tok/s | comfortable |