manuelgscom/dale-0.8b-GGUF
Text Generation • 0.8B • Updated • 308 • 3
dale
Offline natural language → bash for Linux: fine-tuned Qwen3.5, 0.8B on CPU and 2B/4B/9B on GPU (Vulkan), for llama.cpp.
Note 0.8B · the CPU model of the dale CLI (default) · 79.4 % on test v2 · 529 MB
manuelgscom/dale-2b-GGUF
Text Generation • 2B • Updated • 94Note 2B · for llama-server (not used by the dale CLI) · 81.7 % on test v2 · 1.3 GB
manuelgscom/dale-4b-GGUF
Text Generation • 4B • Updated • 79Note 4B · GPU model of the dale CLI (≥ 4 GiB free VRAM) · 87.0 % on test v2 · 2.7 GB
manuelgscom/dale-9b-GGUF
Text Generation • 9B • Updated • 94Note 9B · GPU model of the dale CLI, most accurate (≥ 8 GiB free VRAM) · 89.2 % on test v2 · 5.6 GB
manuelgscom/dale-bash-test-v2
Viewer • Updated • 471 • 52Note Test v2: 471 shell requests (Spanish and English) used to evaluate every dale model