Update: Since accuracy remains our highest priority for an educational model, we will only carry forward the vocabulary-pruning optimization. It improved deployment efficiency and reduced model size without any measured loss in accuracy. Our updated deployment model is therefore Muta-Tutor-Qwen2.5-1.5B-Q4_K_M-vocab32k.gguf, which preserves the capability of our selected Muta Tutor while being smaller and faster. It can be found here.
Log in or sign up for Devpost to join the conversation.