39  Downloads Updated  2 days ago

Compact 1.9B language model in BF16 precision, optimized for MLX workflows. Features thinking/reasoning mode, tool calling, and a massive 262K token context window. Runs on GPUs with 8 GB+ memory.