DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Thinking-EXP2

IMPORTANT: This model has an upgraded Jinja template which repairs issues with org model (repeats, long thinking, loops) and upgrades/repairs to tool handling.

Qwen3.5-21B-Claude-4.6-Opus-Thinking-EXP2

Experimental "Honey I Shrunk the Model" using Qwen 3.5 27B dense as a base, followed by training on Claude 4.6 Opus (small) dataset.

This process removed 16 layers from the model, with training via unsloth to re-stablize/re-train the model.

EXP2 used a different method to "shrink the model".

Not everything may work.

Please report any issues via the community tab.

Thank you.

More experiments in progress.