Offline Spanish proofreader: a fine-tuned 0.8B model (GGUF) and an open benchmark of native-speaker errors.
👋 Open to Work
Manuel Garcia Sanz
manuelgscom
AI & ML interests
AI Architect · Small specialized models that beat 10× larger generalists · Fine-tuning (full, LoRA, SFT, DPO) · Rigorous evaluation · Native C++ inference with llama.cpp (CUDA, Vulkan, CPU) · On-prem & hybrid AI
Recent Activity
updated a collection about 16 hours ago
dale updated a model about 16 hours ago
manuelgscom/dale-0.8b-GGUF updated a model about 16 hours ago
manuelgscom/dale-0.8bOrganizations
None yet
models 13
manuelgscom/dale-0.8b
Text Generation • 0.9B • Updated • 36
manuelgscom/dale-0.8b-GGUF
Text Generation • 0.8B • Updated • 505 • 5
manuelgscom/revisa-0.8b-GGUF
Text Generation • 0.8B • Updated • 171
manuelgscom/factusight-9B-GGUF
Image-Text-to-Text • 9B • Updated • 256 • 1
manuelgscom/factusight-4B-GGUF
Image-Text-to-Text • 4B • Updated • 238 • 1
manuelgscom/factusight-2B-GGUF
Image-Text-to-Text • 2B • Updated • 249 • 1
manuelgscom/askmdb-9b-GGUF
Text Generation • 9B • Updated • 310 • 3
manuelgscom/askmdb-4b-GGUF
Text Generation • 4B • Updated • 214 • 1
manuelgscom/askmdb-2b-GGUF
Text Generation • 2B • Updated • 230 • 1
manuelgscom/askmdb-0.8b-GGUF
Text Generation • 0.8B • Updated • 245 • 1