Launched in July 2024, this smaller model provided high performance with reduced computational costs.