Minos-v1
by NousResearch
396M params · text-classification · 172 likes · 1.9k downloads
Minos-v1 is a 396M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (3317 tok/s), GeForce RTX 5090 (4974 tok/s), M4 Max 128GB (1213 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 3317 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 4974 tok/s | comfortable |
| M4 Max 128GB | 128GB | 1213 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 606 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 606 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 6077 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 12223 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 2932 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 2425 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 2354 tok/s | comfortable |