e5-mistral-7b-instruct
by intfloat
7.1B params · feature-extraction · 560 likes · 209.6k downloads
e5-mistral-7b-instruct is a 7.1B parameter model. At Q4 quantization it requires 4GB of VRAM. It runs comfortably on GeForce RTX 4090 (184 tok/s), GeForce RTX 5090 (276 tok/s), M4 Max 128GB (67 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| SambaNova |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 184 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 276 tok/s | comfortable |
| M4 Max 128GB | 128GB | 67 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 33 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 33 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 338 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 680 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 163 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 135 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 131 tok/s | comfortable |