DGX Spark 本地部署推荐：2 台起可跑 DeepSeek V4.1 Flash 1017
博主给出 DGX Spark 多机部署推荐：单台选 Qwen3.8-Flash-Next，2 台及以上可选 DeepSeek-V4.1-Flash-1017（部分任务更强）或 GLM-5.3-Flash（更稳定），4 台时两者性能都很好。
Here are the BEST models to run on your DGX Sparks Lots of models, lots of combinations you can run, depending on how many sparks you have. But these are MY personal recommendations. ✨ One DGX Spark - have one spark? Don't overthink it. This is THE model to run that will give you the best combination of quality, performance, and KV cache! Qwen3.8-Flash-Next https://mia-ai.net/models/Qwen3.8-Flash-Next-Single-DGX-Spark-TensorFold ✨ 2x DGX Spark - you can't go wrong with either! DeepSeek is better at some tasks, while GLM is more consistent and nicer to talk to. DeepSeek-V4.1-Flash-1017 https://github.com/MiaAI-Lab/DeepSeek-v4.1-Flash-1017-2x-DGX-Sparks GLM-5.3-Flash https://mia-ai.net/models/GLM-5.3-Flash-EXL3-2x-DGX-Sparks-TensorFold ✨ 3x DGX Sparks - same story as above! DeepSeek-V4.1-Flash-1017 https://mia-ai.net/models/DeepSeek-v4.1-Flash-DGX-Sparks GLM-5.3-Flash https://mia-ai.net/models/GLM-5.3-Flash-EXL3-2x-DGX-Sparks-TensorFold ✨ 4x DGX Sparks - super performance on 4 sparks on both of these models! DeepSeek-V4.1-Flash-1017 https://mia-ai.net/models/DeepSeek-v4.1-Flash-DGX-Sparks GLM-5.3-Flash https://mia-ai.net/models/GLM-5.3-Flash-EXL3-2x-DGX-Sparks-TensorFold With 4 Sparks, you can also run a different model on each pair. Personally, I'd only run Qwen3.8-Flash-Next if I had just one DGX Spark. Two or more? Go with GLM or DeepSeek. See all models here: https://mia-ai.net/models