ASUS Ascent GX10 128GB vs NVIDIA DGX Spark 128GB
For running models locally, priced in India. At 32k context the first runs 42 models fully in memory, the second 42.
In short
Both have 128 GB of memory for models; the ASUS Ascent GX10 128GB has 273 GB/s of bandwidth and the NVIDIA DGX Spark 128GB 273 GB/s. At 32k context both run the same 42 models we track fully in memory. On the models both run fully in memory, the ASUS Ascent GX10 128GB and the NVIDIA DGX Spark 128GB are about as fast as each other (estimated).
Updated . Speeds are estimates; how they are worked out.
Specifications
| Spec | ASUS Ascent GX10 128GB | NVIDIA DGX Spark 128GB |
|---|---|---|
| Chip | GB10 Grace Blackwell | GB10 Grace Blackwell |
| Memory bus | 256-bit | 256-bit |
| Memory bandwidth | 273 GB/s | 273 GB/s |
| FP16 compute | 62 TFLOPS | 62 TFLOPS |
| Software | CUDA | CUDA |
| Unified memory | 128 GB | 128 GB |
| GPU can use by default | 119 GB | 119 GB |
| GPU can use, tuned | 122 GB | 122 GB |
| Operating system | dgx-os | dgx-os |
| Processor | 20-core Arm (GB10) | 20-core Arm (GB10) |
| Storage | 1000 GB | 4000 GB |
What each runs
Estimated decode speed with the most precise quant that fits each one est.
| Model | ASUS Ascent GX10 128GB | NVIDIA DGX Spark 128GB |
|---|---|---|
| DeepSeek V4 Flash | 11tok/sUD-Q2_K_XL | 11tok/sUD-Q2_K_XL |
| GLM 5.3 Flash | 11tok/sUD-Q2_K_XL | 11tok/sUD-Q2_K_XL |
| Qwen 3.8 Flash Next | 3tok/sUD-Q4_K_XL | 3tok/sUD-Q4_K_XL |
| MiniMax M2.7 | 8tok/sUD-Q3_K_M | 8tok/sUD-Q3_K_M |
| Qwen 3.8 27B | 5tok/sQ8_0 | 5tok/sQ8_0 |
| Qwen 3.5 122B-A10B | 11tok/sQ6_K | 11tok/sQ6_K |
| Qwen 3.6 27B | 6tok/sQ8_0 | 6tok/sQ8_0 |
| gpt-oss 120B | 23tok/sMXFP4 | 23tok/sMXFP4 |
| Qwen3 Coder Next | 20tok/sQ8_0 | 20tok/sQ8_0 |
| Nemotron 3 Super 120B-A12B | 9tok/sUD-Q6_K | 9tok/sUD-Q6_K |
| Gemma 4 31B | 5tok/sQ8_0 | 5tok/sQ8_0 |
| Mistral Small 4 119B | 15tok/sUD-Q6_K | 15tok/sUD-Q6_K |
| Qwen 3.6 35B-A3B | 27tok/sQ8_0 | 27tok/sQ8_0 |
| Qwen 3.5 27B | 6tok/sQ8_0 | 6tok/sQ8_0 |
| Gemma 4 26B-A4B | 20tok/sQ8_0 | 20tok/sQ8_0 |
| GLM 4.5 Air | 4tok/sQ8_0 | 4tok/sQ8_0 |
| Qwen 3.5 35B-A3B | 27tok/sQ8_0 | 27tok/sQ8_0 |
| Qwen3 Next 80B-A3B | 25tok/sQ8_0 | 25tok/sQ8_0 |
| GLM 4.7 Flash | 14tok/sQ8_0 | 14tok/sQ8_0 |
| Devstral Small 2 24B | 6tok/sQ8_0 | 6tok/sQ8_0 |
| Nemotron 3.5 Lightning 30B-A3B | 28tok/sUD-Q6_K_XL | 28tok/sUD-Q6_K_XL |
| gpt-oss 20B | 31tok/sMXFP4 | 31tok/sMXFP4 |
| Kimi Linear 48B-A3B | 29tok/sQ8_0 | 29tok/sQ8_0 |
| Qwen3 Coder 30B-A3B | 16tok/sQ8_0 | 16tok/sQ8_0 |
| Qwen3 32B | 4tok/sQ8_0 | 4tok/sQ8_0 |
| Gemma 4 12B | 13tok/sQ8_0 | 13tok/sQ8_0 |
| Qwen 3.5 9B | 16tok/sQ8_0 | 16tok/sQ8_0 |
| Mistral Small 3.2 24B | 6tok/sQ8_0 | 6tok/sQ8_0 |
| Gemma 3 27B | 5tok/sQ8_0 | 5tok/sQ8_0 |
| DeepSeek R1 Distill Qwen 32B | 4tok/sQ8_0 | 4tok/sQ8_0 |
| Llama 3.3 70B | 2tok/sQ8_0 | 2tok/sQ8_0 |
| Ministral 3 14B | 9tok/sQ8_0 | 9tok/sQ8_0 |
| Qwen3 14B | 8tok/sQ8_0 | 8tok/sQ8_0 |
| DeepSeek R1 0528 Qwen3 8B | 13tok/sQ8_0 | 13tok/sQ8_0 |
| Qwen3 8B | 13tok/sQ8_0 | 13tok/sQ8_0 |
| Phi-4 14B | 8tok/sQ8_0 | 8tok/sQ8_0 |
| Qwen 3.5 4B | 30tok/sQ8_0 | 30tok/sQ8_0 |
| Ministral 3 8B | 13tok/sQ8_0 | 13tok/sQ8_0 |
| Granite 4.2 8B | 12tok/sQ8_0 | 12tok/sQ8_0 |
| Gemma 4 E4B | 19tok/sQ8_0 | 19tok/sQ8_0 |
| LFM2.5 8B-A1B | 38tok/sQ8_0 | 38tok/sQ8_0 |
| Llama 3.1 8B | 16tok/sQ6_K | 16tok/sQ6_K |
Questions
- Which can run bigger models, the ASUS Ascent GX10 128GB or the NVIDIA DGX Spark 128GB?
- At 32k context both run the same 42 models we track fully in memory.
- Which is faster for local AI, the ASUS Ascent GX10 128GB or the NVIDIA DGX Spark 128GB?
- On the models both run fully in memory, the ASUS Ascent GX10 128GB and the NVIDIA DGX Spark 128GB are about as fast as each other (estimated).