4× RTX PRO 6000
189tok/sPeak
1 request, context 0 (78 prompt tokens), synthetic sustained decode with ignore-EOS, with MTP3 (per recipe) Source
Intelligence42with thinking
- Custom kernel required
Local AI · Creator
brandonmusic
6 published recipes.
Recipes
2× RTX PRO 6000
152tok/sPeak
1 request, best value reported by source, context 64k Source
Intelligence42with thinking
- ≤ 3 bit: quantization may cost quality
- Non-commercial
- License unclear
4× RTX PRO 6000
101tok/sPeak
1 request, best value reported by source Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
4× RTX PRO 6000
74tok/sPeak
1 request, best value reported by source Source
Intelligence34with thinkingno thinking 22
4× RTX PRO 6000
62.6tok/sPeak
1 request, best value reported by source, context 128k Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
4× RTX PRO 6000
48.9tok/sPeak
1 request, best value reported by source Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality