Introducing the Palmyra-mini family: Powerful, lightweight, and ready to reason!

The Palmyra-mini family of models has been introduced, featuring a lightweight architecture designed for efficient reasoning tasks. These models range in size from 125M to 1B parameters and achieve state-of-the-art performance on various reasoning benchmarks, demonstrating significant improvements in accuracy and inference speed compared to previous iterations. This release is particularly relevant for practitioners seeking efficient LLMs that maintain high performance while minimizing resource consumption in deployment scenarios.