Cortex-X925
NVIDIA NVIDIA_DGX_Spark
122 GB RAM · 14 models tested
Avg Score
77 /100
Avg Speed
33.6 tok/s
Models Tested
14
| Model | Size | Score | tok/s | TTFT | Verdict |
|---|---|---|---|---|---|
| gpt-oss:20b | 20.9B MXFP4 | 93 | 60.5 | 3.1s | Excellent |
| gpt-oss:20b | 20.9B MXFP4 | 92 | 60.7 | 3.2s | Excellent |
| gemma4:12b | 11.9B Q4_K_M | 86 | 26.3 | 545 ms | Excellent |
| qwen3:30b-a3b | 30.5B Q4_K_M | 84 | 92.7 | 1.5s | Excellent |
| deepseek-coder-v2:16b | 15.7B Q4_0 | 82 | 117.2 | 275 ms | Excellent |
| qwen3:32b | 32.8B Q4_K_M | 79 | 9.7 | 324 ms | Good |
| qwen2.5-coder:32b | 32.8B Q4_K_M | 78 | 9.5 | 395 ms | Good |
| phi4:14b | 14.7B Q4_K_M | 77 | 22.9 | 198 ms | Good |
| glm-4.7-flash:bf16 | 29.9B F16 | 76 | 25.8 | 605 ms | Good |
| mistral-small:24b | 23.6B Q4_K_M | 75 | 14.3 | 327 ms | Good |
| qwen2.5:72b | 72.7B Q4_K_M | 75 | 4.4 | 529 ms | Not Rec. |
| llama3.3:70b | 70.6B Q4_K_M | 73 | 4.6 | 560 ms | Not Rec. |
| codestral:22b | 22.2B Q4_0 | 67 | 17.2 | 129 ms | Good |
| deepseek-r1:70b | 70.6B Q4_K_M | 47 | 4.7 | 557 ms | Not Rec. |