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 · 4× RTX PRO 6000
Kimi-K2.6 519B NVFP4 with vLLM on 4× RTX PRO 6000
Kimi-K2.6 519B NVFP4 with vLLM on 4× RTX PRO 6000: 71 tok/s according to huggingface.co (dataset as of Nov 3, 2026).
by 0xSero
Intelligence (original model) · Artificial Analysisno independent value
Engine
- Engine
- vLLM
- Quantization
- NVFP4, fp8_e4m3 KV
- Model family
- Kimi-K2.6
- Context
- 262,144
- Parameters
- 519B (192 von 384 experts)
- Creator
- 0xSero
- GitHub stars
- 7
- Repo updated
- Sep 4, 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
Time to first token
Shorter is better.
What you need
- Hardware
- 4 × NVIDIA RTX PRO 6000 Blackwell
- Weights
- Main weights0xSero/Kimi-K2.6-519B-NVFP4 other
- Basenvidia/Kimi-K2.6-NVFP4 other
- Draft modellightseekorg/kimi-k2.6-eagle3-mla other
- Engine
- vLLM
- Context
- 262,144 tokens
Notes
What matters before you rebuild it.
- REAP-pruned
- License unclear
Sources
Related recipes
4× RTX PRO 6000
139tok/sEveryday
1 request, prose prompt, with MTP (per recipe) Source
Intelligence18with thinkingno thinking 21
- Custom kernel required
4× RTX PRO 6000
81.7tok/sEveryday
1 request, prose prompt, with MTP Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
4× RTX PRO 6000
73.1tok/sEveryday
1 request, prose prompt, context 32788, with DSPARK Source
Intelligence40with thinkingno thinking 25
4× RTX PRO 6000
70.7tok/sEveryday
1 request, prose prompt, with MTP Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
4× RTX PRO 6000
64tok/sEveryday
1 request, prose prompt, with MTP (per recipe) Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
- Custom kernel required
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.