DSpark drafter heads for llama.cpp (measured speedups) and the self-distillation datasets behind them.
Sat Grewal PRO
satgeze
AI & ML interests
Optimise AI models for consumer grade systems resources
Recent Activity
published a model about 7 hours ago
satgeze/DeepSeek-v4.1-Flash-EXL3-2.0bpw-Ablit-EngramQ4-SM120-Dual-RTX-Pro-6000 liked a model about 7 hours ago
diffbot/DeepSeek-V4.1-Flash-EXL3-2.0bpw-2x-RTX-PRO-6000 liked a model 4 days ago
dealignai/DeepSeek-V4.1-Flash-UNCENSORED-FP8Organizations
None yet
Uncensored trunks with YaRN 1M context, needle-certified, plus MTP speculative decoding and vision. Full archives on ModelScope.
-
satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 0.4B • Updated • 2.76k • 9 -
satgeze/Gemma4-12B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 12B • Updated • 2.76k • 6 -
satgeze/Gemma4-26B-A4B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 25B • Updated • 1.74k • 6 -
satgeze/Gemma4-31B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 31B • Updated • 890 • 3
DSpark drafter heads for llama.cpp (measured speedups) and the self-distillation datasets behind them.
Models needle-certified past one million tokens, plus the ceiling studies. All raw data published, misses included.
Uncensored trunks with YaRN 1M context, needle-certified, plus MTP speculative decoding and vision. Full archives on ModelScope.
-
satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 0.4B • Updated • 2.76k • 9 -
satgeze/Gemma4-12B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 12B • Updated • 2.76k • 6 -
satgeze/Gemma4-26B-A4B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 25B • Updated • 1.74k • 6 -
satgeze/Gemma4-31B-Uncensored-HauhauCS-1M-GGUF
Text Generation • 31B • Updated • 890 • 3
Ornith-1.0 with YaRN 1M context, needle-certified, MTP speculative decoding and vision. Full quant ladders on ModelScope.