harrier-oss-v1-0.6b
by microsoft
596M params · feature-extraction · 257 likes · 307.4k downloads
harrier-oss-v1-0.6b is a 596M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (2202 tok/s), GeForce RTX 5090 (3303 tok/s), M4 Max 128GB (806 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| HF Inference |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 2202 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 3303 tok/s | comfortable |
| M4 Max 128GB | 128GB | 806 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 403 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 403 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 4035 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 8117 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 1947 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 1610 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 1563 tok/s | comfortable |