QRWKV6-32B-Instruct-Preview-v0.1
by recursal
34.7B params · text-generation · 79 likes · 6 downloads
QRWKV6-32B-Instruct-Preview-v0.1 is a 34.7B parameter model. At Q4 quantization it requires 17GB of VRAM. It runs comfortably on GeForce RTX 4090 (37 tok/s), GeForce RTX 5090 (56 tok/s), A100 PCIe 80 GB (69 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| Featherless |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 37 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 56 tok/s | comfortable |
| M4 Max 128GB | 128GB | 13 tok/s | tight |
| M4 Pro 48GB | 48GB | 6 tok/s | tight |
| M4 Pro 24GB | 24GB | 6 tok/s | tight |
| A100 PCIe 80 GB | 80GB | 69 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 139 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 33 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 27 tok/s | tight |
| M4 Max 64GB | 64GB | 13 tok/s | tight |