llama-claude-reasoning

Llama-3.2-3B-claude-reasoning-GGUF : GGUF

Trained with: Unsloth Studio

This model was finetuned and converted to GGUF format using Unsloth.

Usecases

From my testing, the only good usecase is to mimic the style of output and reasoning of larger models like claude 4.5 opus. Beyond that, it keeps llama's "yes-man" style and has no clue what it is doing. Maybe it got a bit better but the fact its 3 billion parameters limits its ability. also it sometimes forgets to close thinking tags.

Benchmarks

coming soon to see if its any good

Datasets:

TeichAI/claude-4.5-opus-high-reasoning-250x

Example usage:

• For text only LLMs: ./llama.cpp/llama-cli -hf Aadeshisdoingsomething/llama-claude-reasoning --jinja
• For multimodal models: ./llama.cpp/llama-mtmd-cli -hf Aadeshisdoingsomething/llama-claude-reasoning --jinja

Available Model files:

• llama-3.2-3b-instruct.Q4_K_M.gguf

Unsloth

This was trained 2x faster with Unsloth <img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>