What can the Dual DGX Spark run?
| Text models: 81 | Visual models: 0 | Strongest first |
- Memory
- 256 GB unified
- Bandwidth
- 273 GB/s each
- Maker
- NVIDIA
- Also called
- 2× DGX Spark
Open the calculator on this machineDGX Spark (GB10) in our inventory
Text models that run on it
| Model | Best fit | Context | Speed | Strongest at |
|---|---|---|---|---|
| MiMo V2.6 Pro | 1.6-bit | 608K | 20 tok/s | Chat, Coding, Agents, Docs |
| DeepSeek V4.1 Flash | 1.6-bit | 1M | 48 tok/s | Chat, Coding, Agents, Docs |
| GLM-5.3 | 1.6-bit | 864K | 21 tok/s | Chat, Coding, Agents, Docs |
| GLM-5.3 Flash | 5-bit | 1M | 14 tok/s | Chat, Coding, Agents, Docs |
| Qwen3.8 Flash-Next | 8-bit | 256K | 27 tok/s | Chat, Coding, Agents |
| Muse Spark Open 816B A42B | 1.6-bit | 1M | 20 tok/s | Chat, Docs |
| MiMo V2.6 Flash | FP4 + FP8 | 1M | 17 tok/s | Coding, Agents |
| MiniMax M3 | 3-bit | 256K | 16 tok/s | Docs |
| DeepSeek V4 Flash | FP4 + FP8 | 1M | 9.9 tok/s | Docs |
| Kimi K2.6 | 1.6-bit | 256K | 26 tok/s | Chat, Coding, Agents, Docs |
| Hy3 | 5-bit | 96K | 12 tok/s | Chat, Coding, Agents, Docs |
| Nemotron 3 Ultra 550B A55B | 2-bit | 256K | 8.6 tok/s | Chat, Coding, Docs |
| Qwen3.5 397B A17B | 3-bit | 256K | 21 tok/s | Agents, Docs |
| MiniMax M2.7 | FP8 | 52K | 18 tok/s | Coding, Agents |
| DeepSeek V3.2 | 1.6-bit | 160K | 23 tok/s | Chat, Coding, Agents, Docs |
| Mistral Medium 3.5 128B | FP8 | 256K | 1.4 tok/s | Chat, Coding, Agents, Docs |
| Qwen3.5 122B A10B | 8-bit | 256K | 17 tok/s | Chat, Coding, Agents, Docs |
| Nemotron 3 Super 120B A12B | 8-bit | 256K | 14 tok/s | Chat, Coding, Agents, Docs |
| Nemotron 3.5 Bolt 74B A9B | 8-bit | 256K | 18 tok/s | Chat, Coding, Agents, Docs |
| Mistral Small 4 119B | FP8 | 256K | 27 tok/s | Chat, Docs |
| gpt-oss 120B | MXFP4 | 128K | 47 tok/s | Chat, Coding |
| Qwen3 Coder Next 80B | 8-bit | 256K | 50 tok/s | Coding, Agents |
| Llama 3.3 70B | 8-bit | 128K | 2.6 tok/s | Chat, Coding, Agents, Docs |
| Inkling | 1.6-bit | 4K | 21 tok/s | Text and images |
| DeepSeek R1 | 1.6-bit | 160K | 21 tok/s | Not rated |
| DeepSeek V3 | 1.6-bit | 160K | 21 tok/s | Not rated |
| GLM 5 | 1.6-bit | 198K | 20 tok/s | Not rated |
| GLM 4.5 Air | 8-bit | 128K | 12 tok/s | Not rated |
| Inkling Small | 6-bit | 4K | 19 tok/s | Text and images |
| MiniMax M2.5 | FP8 | 52K | 15 tok/s | Not rated |
| GLM 5.1 | 1.6-bit | 198K | 20 tok/s | Not rated |
| Laguna M.1 | 6-bit | 200K | 9.5 tok/s | Not rated |
| Laguna S 2.1 | 8-bit | 1M | 20 tok/s | Not rated |
| Qwen3 VL 235B A22B Instruct | 6-bit | 256K | 10.0 tok/s | Text and images |
| GLM 5.2 | 1.6-bit | 864K | 20 tok/s | Not rated |
| Qwen3 235B A22B | 6-bit | 40K | 10.0 tok/s | Not rated |
| Ornith 1.5 122B | 8-bit | 256K | 13 tok/s | Not rated |
| Kimi K2 Instruct | 1.6-bit | 128K | 21 tok/s | Not rated |
| Ornith 1.5 397B | 3-bit | 256K | 15 tok/s | Not rated |
| Ornith 1.0 397B | 4-bit | 96K | 16 tok/s | Not rated |
| Step 3.5 Flash | 8-bit | 72K | 18 tok/s | Not rated |
| DeepSeek V3 0324 | 1.6-bit | 160K | 21 tok/s | Not rated |
| Qwen2.5 72B Instruct | 8-bit | 32K | 2.5 tok/s | Not rated |
| Qwen3 Next 80B A3B Instruct | 8-bit | 256K | 50 tok/s | Not rated |
| DeepSeek R1 0528 | 1.6-bit | 160K | 21 tok/s | Not rated |
| GLM 4.5 | 4-bit | 64K | 8.4 tok/s | Not rated |
| Llama 4 Scout 17B 16E Instruct | 8-bit | 608K | 9.7 tok/s | Text and images |
| MiniMax M2 | FP8 | 52K | 15 tok/s | Not rated |
| MiMo V2.5 | 5-bit | 960K | 15 tok/s | Not rated |
| Qwen2.5 VL 72B Instruct | 8-bit | 125K | 2.5 tok/s | Text and images |
| Solar Open 100B | 8-bit | 128K | 14 tok/s | Not rated |
| GLM 5.3 Flash BF16 | 5-bit | 1M | 14 tok/s | Text and images |
| Qwen3 VL 235B A22B Thinking | 6-bit | 256K | 10.0 tok/s | Text and images |
| Qwen3 235B A22B Instruct 2507 | 6-bit | 256K | 10.0 tok/s | Not rated |
| NVIDIA Nemotron 3.5 Super VL 120B A12B BF16 | 8-bit | 1M | 14 tok/s | Text and images |
| Hy3 Preview | 5-bit | 88K | 12 tok/s | Not rated |
| Qwen3 Coder 480B A35B Instruct | 3-bit | 36K | 11 tok/s | Coding |
| GLM 4.7 | 4-bit | 64K | 8.4 tok/s | Not rated |
| Qwen3 Next 80B A3B Thinking | 8-bit | 256K | 50 tok/s | Not rated |
| GLM 5.3 BF16 | 1.6-bit | 864K | 20 tok/s | Not rated |
Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 81 are shown.
Questions
What is the best LLM for the Dual DGX Spark?
Of the open models we track, the strongest that run well on the Dual DGX Spark are MiMo V2.6 Pro, DeepSeek V4.1 Flash, GLM-5.3. 81 text models fit it in all.
Calculated from published specs, not benchmarked. All machines