4× RTX PRO 6000
113tok/sPeak
1 request, best value reported by source Source
Intelligence29with thinking
- License unclear
Local AI · Recipe · 4× DGX Spark
MiniMax-M3 NVFP4 with vLLM on 4× DGX Spark
MiniMax-M3 NVFP4 with vLLM on 4× DGX Spark: 33.7 tok/s according to github.com (dataset as of Nov 3, 2026).
by tonyd2wild
Engine
- Engine
- vLLM
- Quantization
- AWQ INT4 weights + NVFP4 KV
- Model family
- MiniMax-M3
- Context
- 1,048,576
- Creator
- tonyd2wild
- GitHub stars
- 1
- Repo updated
- Jul 13, 2026
Measurements
Every sourced value of this recipe, each with its condition and source. Bars relative to the largest value in the group.
Decode · 1 request
What you need
- Hardware
- 4 × NVIDIA DGX Spark (GB10)
- Weights
- Main weightscyankiwi/MiniMax-M3-AWQ-INT4 other
- Draft modelInferact/MiniMax-M3-EAGLE3 MIT
- Engine
- vLLM
- Context
- 1,048,576 tokens
Notes
What matters before you rebuild it.
- Experimental
- Custom kernel required
- License unclear
Sources
Related recipes
4× RTX PRO 6000
87tok/sPeak
1 request, best value reported by source Source
Intelligence29with thinking
- License unclear
4× RTX PRO 6000
39.2tok/sPeak
1 request, synthetic prompt Source
Intelligence29with thinking
- License unclear
4× DGX Spark
38tok/sPeak
1 request, best value reported by source Source
Intelligence29with thinking
- License unclear
2× DGX Spark
36.6tok/sPeak
1 request, JSON prompt, context 196k Source
Intelligence29with thinking
- License unclear
3× DGX Spark
10.5tok/sPeak
1 request, best value reported by source Source
Intelligence29with thinking
- License unclear
Local AI in your company?
In a workshop we work out which models and which hardware fit your tasks, and build the first agent on your infrastructure.