▸ DEVICE UNDER TEST
AMD Radeon Instinct MI325X — 288 GB VRAM.
▸ RADEON INSTINCT MI325X SPEC
- BRAND
- AMD
- VRAM
- 288 GB HBM3e
- BANDWIDTH
- 10300 GB/s
- FP16 COMPUTE
- 653.7 TFLOPS
- FP32 COMPUTE
- 81.7 TFLOPS
- STREAM PROCESSORS
- 19,456
- TDP
- 1000 W
- ARCHITECTURE
- CDNA 3.0
- MSRP
- $20000
▸ AI CAPABILITY
425/ 449 models @ Q4
With 288 GB VRAM and 10300 GB/s bandwidth, this GPU handles models up to 405B parameters.
Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~1177 tok/s.
§ 01TOP MODELS FOR RADEON INSTINCT MI325X
425 FIT · SHOWING 20| MODEL | SIZE | VRAM Q4 | TOK/S | AVG |
|---|---|---|---|---|
| Llama 3.1 405B | 405B | 248.0 GB | 23 | 48.4 |
| Hermes 4 405B | 405B | 248.0 GB | 23 | 33.4 |
| Cogito v2 405B | 405B | 248.0 GB | 23 | — |
| Llama-4-Maverick-17B-128E | 400B | 245.0 GB | 539 | 46.1 |
| Jamba Large 1.6 | 398B | 243.8 GB | 97 | 11.0 |
| Qwen3.5-397B-A17B | 397B | 243.2 GB | 539 | 68.4 |
| Hunyuan-Large | 389B | 238.3 GB | 176 | 73.9 |
| GLM 4.7 | 358B | 219.3 GB | 286 | 64.4 |
| GLM 4.6 | 357B | 218.7 GB | 763 | 41.4 |
| GLM 4.5 | 355B | 217.5 GB | 763 | 47.5 |
| Nemotron-4 340B | 340B | 208.3 GB | 27 | 65.4 |
| MiMo V2.5 | 310.78B | 190.5 GB | 610 | 57.5 |
| MiMo V2 Flash 309B | 309B | 189.4 GB | 610 | 72.4 |
| DeepSeek V4 Flash 0731 | 304.18B | 186.4 GB | 704 | 62.4 |
| ERNIE-4.5-300B-A47B-Paddle | 300.5B | 184.2 GB | 30 | 37.4 |
| ERNIE 4.5 300B-A47B | 300B | 183.9 GB | 195 | 49.8 |
| Hunyuan Hy3-preview | 295B | 180.8 GB | 436 | 62.7 |
| DeepSeek V4 Flash | 290.94B | 178.3 GB | 704 | 59.3 |
| DeepSeek V2 236B | 236B | 144.7 GB | 436 | 40.4 |
| DeepSeek-V2.5 236B | 236B | 144.7 GB | 436 | 64.0 |