4× RTX PRO 6000
235tok/sEveryday
1 request, prose prompt, context 19, with DSpark (per recipe) Source
Intelligence34with thinking
- Custom kernel required
Local AI · Recipe · 2× RTX 3090
DeepSeek-V4-Flash 284B with llama.cpp on 2× RTX 3090
DeepSeek-V4-Flash 284B with llama.cpp on 2× RTX 3090: 21.9 tok/s according to github.com (dataset as of Nov 3, 2026).
by noonghunna
Engine
- Engine
- llama.cpp (club3090 build)
- Quantization
- GGUF Q8-moecache, Experten im RAM
- Model family
- DeepSeek-V4-Flash
- Context
- 200,000
- Parameters
- 284B MoE
- Creator
- noonghunna
- GitHub stars
- 2,471
- Repo updated
- Nov 2, 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
Prefill by context
What you need
- Hardware
- 2 × NVIDIA RTX 3090
- Weights
- Main weightsunsloth/DeepSeek-V4-Flash-0731-GGUF MIT
- Engine
- llama.cpp (club3090 build)
- Context
- 200,000 tokens
Notes
What matters before you rebuild it.
- Partly in system RAM
- Experimental
Sources
Related recipes
4× DGX Spark
46tok/sEveryday
1 request, prose prompt, with DSpark Source
Intelligenceno independent value
4× DGX Spark
42tok/sEveryday
1 request, prose prompt, with DSpark (per recipe) Source
Intelligenceno independent value
- Custom kernel required
3× DGX Spark
41.9tok/sEveryday
1 request, prose prompt, with DSpark (per recipe) Source
Intelligenceno independent value
- Custom kernel required
2× DGX Spark
33.2tok/sEveryday
1 request, prose prompt, with DSpark (per recipe) Source
Intelligenceno independent value
- Custom kernel required
1× DGX Spark
30.2tok/sEveryday
1 request, story prompt, prompt 18 tokens, with MTP (per recipe) Source
Intelligence34with thinking
- ≤ 3 bit: quantization may cost quality
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.