During February 2026, work on the BerriAI/litellm repository centered on integrating Alibaba Cloud’s Qwen3-Max model to deliver enterprise-ready model support and scalable inference. The feature introduced tiered pricing and enhanced capabilities, enabling cost-aware usage and support for large-context inference with up to 258K input and 65K output tokens. The implementation leveraged API integration and backend development skills, utilizing JSON for configuration and data handling. Function calling and tool choice support were added to improve model interactivity, while enterprise parameters and reasoning capabilities were enabled to address diverse use cases and lay the foundation for cloud-optimized deployments.