Romanian-first bilingual (RO/EN), dual-mode reasoner models
AI & ML interests
Insanely fast LLM pre-training and fine-tuning for modern NVIDIA GPUs.
Recent Activity
models 39
surogate/Surogate-3.5-4B
Text Generation • 5B • Updated • 989 • 2
surogate/Surogate-3.5-2B
Text Generation • 3B • Updated • 943 • 1
surogate/Qwen3.5-4B-FP8
Image-Text-to-Text • 5B • Updated • 43
surogate/Qwen3-1.7B-Libra-MF
Text Generation • 2B • Updated • 82
surogate/Qwen3.5-2B-Libra-YTD
Text Generation • 2B • Updated • 582
surogate/Qwen3.5-9B-NVFP4
Image-Text-to-Text • 7B • Updated • 883 • 1
surogate/Qwen3.5-4B-NVFP4
Image-Text-to-Text • 3B • Updated • 158 • 1
surogate/Qwen3.5-2B-NVFP4
Image-Text-to-Text • 2B • Updated • 3
surogate/Qwen3.5-0.8B-NVFP4
Image-Text-to-Text • 0.7B • Updated • 46
surogate/Qwen3.5-9B-FP8
Image-Text-to-Text • 9B • Updated • 5.9k
datasets 11
surogate/hellaswag-ro
Viewer • Updated • 9.25k • 14
surogate/cc-pretrain
Viewer • Updated • 981 • 11
surogate/brd-en
Viewer • Updated • 143 • 8
surogate/brd
Viewer • Updated • 143 • 11
surogate/densemax-self-cognition
Viewer • Updated • 124 • 9
surogate/self-cognition-dan
Viewer • Updated • 2k • 8
surogate/self-cognition-generated
Viewer • Updated • 2k • 14
surogate/self-cognition-qwen3
Viewer • Updated • 50 • 12
surogate/self-cognition
Viewer • Updated • 50 • 6
surogate/alpaca-gpt4-data-en
Viewer • Updated • 52k • 17