Distilled reasoning model based on Qwen 2.5 32B using DeepSeek R1 outputs.
Add a second model to compare.
This model hasn't been benchmarked yet.