← Every GPUNVIDIA · 128 GB unified memory · 273 GB/s

What can the NVIDIA DGX Spark run?

With 128 GB, 10 of 12 popular open models fit. The best pick is gpt-oss 120B at about 33 tok/s, faster than you can read.

ollama run gpt-oss:120b

First time? Install Ollama, then paste that into a terminal. Is this your machine? Check it on the homepage.

128 GB unified memory · about 96 GB for models

Every model the NVIDIA DGX Spark can run

  • Qwen3 4BQuick everyday chat and summaries
    ~60 tok/s
    Runs greatqwen3:4b
  • Qwen3 8BA solid all-rounder for chat and writing
    ~29 tok/s
    Usableqwen3:8b
  • Gemma 3 12BReads images as well as text
    ~19 tok/s
    Usablegemma3:12b
  • Qwen3 14BNoticeably smarter at reasoning and code
    ~16 tok/s
    Usableqwen3:14b
  • gpt-oss 20BOpenAI's open model. Fast, strong reasoning
    ~48 tok/s
    Runs greatgpt-oss:20b
  • Gemma 3 27BGoogle's best open model for its size
    ~9 tok/s
    Slowgemma3:27b
  • Qwen3 30B-A3B30B smarts at small-model speed
    ~48 tok/s
    Runs greatqwen3:30b
  • Qwen3 32BStrong coding and long answers
    ~8 tok/s
    Slowqwen3:32b
  • Llama 3.3 70BClose to the big cloud models
    ~3 tok/s
    Slowllama3.3:70b
  • gpt-oss 120BOpenAI's largest open model
    ~33 tok/s
    Runs greatgpt-oss:120b
  • Qwen3 235B-A22BAlibaba's flagship open model, for multi-GPU rigs
    needs 144 GB
    Too bigqwen3:235b
  • DeepSeek R1 671BThe full DeepSeek R1, for the biggest rigs
    needs 406 GB
    Too bigdeepseek-r1:671b
Common questions

Can the NVIDIA DGX Spark run it?

Can the NVIDIA DGX Spark run Llama 3.3 70B?

Yes. Llama 3.3 70B needs about 45 GB and the NVIDIA DGX Spark with 128 GB of unified memory leaves about 96 GB for models. Expect about 3 tok/s, slow, but it works.

Can the NVIDIA DGX Spark run gpt-oss 120B?

Yes. gpt-oss 120B needs about 67 GB and the NVIDIA DGX Spark with 128 GB of unified memory leaves about 96 GB for models. Expect about 33 tok/s, faster than you can read.

Can the NVIDIA DGX Spark run Qwen3 235B-A22B?

No. Qwen3 235B-A22B needs about 144 GB and the NVIDIA DGX Spark has at most 128 GB of unified memory, of which about 96 GB is usable for models. The most capable model it runs well is gpt-oss 120B, at about 33 tok/s.

How we estimate

Generating a token reads the model's active weights from memory once, so speed is roughly memory bandwidth (273 GB/s) divided by the active weight size, with efficiency calibrated on our own rack. Sizes are Ollama's default 4-bit downloads plus about 1.2 GB for context. The GPU shares memory with the system, so we count about three quarters of it for models. Real results are usually within about a third. Measured one yourself? Send us the number.

Want something bigger?

The Dell DSS 8440 + 10× Tesla P100 runs Qwen3 235B-A22B at about 19 tok/s, from $5,100 used.

See the builds

Compare

Similar cards

Every GPU we track

Head to head

The NVIDIA DGX Spark against its rivals

Every comparison