DGX Spark / GB10
HP ZGX Nano G1n
4TB
128 GB of LPDDR5x unified at 273 GB/s. Runs up to Qwen3.5 122B-A10B, Nemotron 3 Super 120B-A12B, GPT-OSS 120B.
- Memory
- 128 GB ~109 GB usable for models
- Bandwidth
- 273 GB/s sets tokens/sec
- Software
- CUDA CUDA
- Price per GB
- $60.15 of model memory
Where to buy
- B&H Photo $7,699Out of stockCheck price at B&H
What it runs
Estimated fit and decode speed for one user at ~8K context. How we estimate
- Qwen3.5 9Bdense · 9B active
Q4_K_MFits~30 tok/s7 of 109 GBQ8_0Fits~17 tok/s12 of 109 GB
- Gemma 4 12Bdense · 12B active
Q4_K_MFits~23 tok/s9 of 109 GBQ8_0Fits~13 tok/s15 of 109 GB
- GPT-OSS 20BMoE · 3.6B active
MXFP4 (native)Fits~86 tok/s13 of 109 GBQ8_0Fits~43 tok/s25 of 109 GB
- Gemma 4 26B-A4BMoE · 3.8B active
Q4_K_MFits~71 tok/s18 of 109 GBQ8_0Fits~41 tok/s30 of 109 GB
- Qwen3.8 27Bdense · 27B active
Q4_K_MFits~10 tok/s19 of 109 GBQ8_0Fits~6 tok/s32 of 109 GB
- GLM-4.7-Flash 30B-A3BMoE · 3B active
Q4_K_MFits~90 tok/s21 of 109 GBQ8_0Fits~51 tok/s35 of 109 GB
- Gemma 4 31Bdense · 31B active
Q4_K_MFits~9 tok/s21 of 109 GBQ8_0Fits~5 tok/s36 of 109 GB
- Qwen3.6 35B-A3BMoE · 3B active
Q4_K_MFits~90 tok/s24 of 109 GBQ8_0Fits~51 tok/s41 of 109 GB
- Llama 3.3 70Bdense · 70B active
Q4_K_MFits~4 tok/s46 of 109 GBQ8_0Fits~2 tok/s80 of 109 GB
- Qwen3-Next 80B-A3BMoE · 3B active
Q4_K_MFits~90 tok/s52 of 109 GBQ8_0Fits~51 tok/s91 of 109 GB
- Llama 4 Scout 109B-A17BMoE · 17B active
Q4_K_MFits~16 tok/s71 of 109 GBQ8_0Won't fitneeds ~123 GB
- GPT-OSS 120BMoE · 5.1B active
MXFP4 (native)Fits~60 tok/s67 of 109 GBQ8_0Won't fitneeds ~132 GB
- Nemotron 3 Super 120B-A12BMoE · 12B active
Q4_K_MFits~23 tok/s78 of 109 GBQ8_0Won't fitneeds ~135 GB
- Qwen3.5 122B-A10BMoE · 10B active
Q4_K_MFits~27 tok/s79 of 109 GBQ8_0Won't fitneeds ~138 GB
- Qwen3 235B-A22BMoE · 22B active
Q4_K_MWon't fitneeds ~151 GBQ8_0Won't fitneeds ~264 GB
- Muse Spark Open 816B-A42BMoE · 42B active
Q4_K_MWon't fitneeds ~521 GBQ8_0Won't fitneeds ~912 GB
Image: HP product page og:image