Phi-4 Finetuning + Bug Fixes by UnslothJan 10, 2025 • By Daniel & MichaelJan 10, 2025•By Daniel & MichaelPhi-4, Microsoft's new 14B model that performs on par with OpenAI's GPT-4o-mini is now in Unsloth! We found & fixed 4 bugs in Phi-4, greatly increasing the model’s accuracy—details below.Unsloth makes Phi-4 finetuning 2x faster, use 70% less memory, and enables >128K context lengths which is 12x longer than Hugging Face + FA2’s 12K on a 48GB GPU.We converted Phi-4 to Llama’s architecture for ...