Model Registry

Microsoft's state-of-the-art 3.8B parameter model, optimized by Unsloth for exceptionally high performance in a compact footprint.

Next-generation lightweight Qwen model with Ultra-Discrete quantization for superior performance at 0.8B scale.

IBM's ultra-lightweight Mamba-2/Transformer hybrid model, highly optimized by Unsloth for low-latency tool-calling and edge execution.

A fine-tuned 3B parameter model, optimized by Unsloth for high performance and quantized with Ultra-Discrete IQ1_S to run on low-resource edge nodes.