Collections

Qwen3.8-Flash-Next GGUFs with the PLE (n-gram table) split into ONE shared file for all quants. From Unsloth; llama.cpp needs the --ple-gguf patch.

CorX language models, built from scratch in Jamaica: from the 160M CorX1.5 to the 27B Patois-speaking CorX3.8.

A family of task-specific models built on Aster base models, combining specialized training techniques, additional heads, and processing pipelines.

Curated by Brandon Brooks (Bright). Swedish ASR runtimes. Hardware-specific benchmarks; web demos may upload audio. Not independently tested.