AutoGLM-Phone-9B-Multilingual
by zai-org
934K params · image-text-to-text · 229 likes · 14.1k downloads
AutoGLM-Phone-9B-Multilingual is a 934K parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (1405179 tok/s), GeForce RTX 5090 (2107234 tok/s), M4 Max 128GB (514212 tok/s).
Inference providers
| Provider | $/1M in | $/1M out | Throughput |
|---|---|---|---|
| Novita | 62 tok/s |
GPU compatibility
| GPU | VRAM | Q4 Decode | Verdict |
|---|---|---|---|
| GeForce RTX 4090 | 24GB | 1405179 tok/s | comfortable |
| GeForce RTX 5090 | 32GB | 2107234 tok/s | comfortable |
| M4 Max 128GB | 128GB | 514212 tok/s | comfortable |
| M4 Pro 48GB | 48GB | 257106 tok/s | comfortable |
| M4 Pro 24GB | 24GB | 257106 tok/s | comfortable |
| A100 PCIe 80 GB | 80GB | 2574486 tok/s | comfortable |
| H100 SXM5 80 GB | 80GB | 5178082 tok/s | comfortable |
| GeForce RTX 3090 | 24GB | 1242388 tok/s | comfortable |
| Radeon RX 7900 XTX | 24GB | 1027397 tok/s | comfortable |
| GeForce RTX 4080 | 16GB | 997260 tok/s | comfortable |