Zyphra Releases Zamba2-1.2B-Instruct and Zamba2-2.7B-Instruct: A New State-of-the-Art Small Language Model Series that Outperforms Gemma2-2B-Instruct

Zyphra has unveiled two new AI language models: Zamba2-1.2B-Instruct and Zamba2-2.7B-Instruct. These models feature a hybrid architecture combining state-space and transformer elements, offering enhanced multi-turn chat capabilities and instruction-following abilities while maintaining efficiency.

The models outperform larger competitors in benchmarks like MT-Bench and IFEval, with Zamba2-2.7B-Instruct scoring 72.40 and 48.02 respectively. Their innovative design allows for faster generation times and lower latency, making them suitable for real-time applications and resource-constrained environments.