manuelgscom/askmdb-0.8b-GGUF
Text Generation • 0.8B • Updated • 60
askmdb
Natural language → MongoDB aggregation pipelines, offline: fine-tuned Qwen3.5 GGUFs. CLI: 0.8B on CPU, 4B/9B on GPU (Vulkan); 2B for llama-server.
Note 0.8B · CPU model of the askmdb CLI (default) · 83.6 % on the test · 529 MB
manuelgscom/askmdb-2b-GGUF
Text Generation • 2B • Updated • 49Note 2B · for llama-server (not used by the askmdb CLI) · 86.8 % on the test · 1.3 GB
manuelgscom/askmdb-4b-GGUF
Text Generation • 4B • Updated • 52Note 4B · GPU model of the askmdb CLI (≥ 4 GiB free VRAM) · 89.6 % on the test · 2.7 GB
manuelgscom/askmdb-9b-GGUF
Text Generation • 9B • Updated • 49 • 2Note 9B · GPU model of the askmdb CLI, most accurate (≥ 8 GiB free VRAM) · 91.6 % on the test · 5.6 GB
manuelgscom/askmdb-test
Viewer • Updated • 500 • 89Note Test set: 500 questions over 10 synthetic MongoDB databases, scored by execution (CC BY 4.0)