DigiNews

Summary

vLLM v0.28.0 release notes reveal major performance improvements (Kimi-K3, DCP, DeepSeek V4), new defaults, and extensive model/hardware support across multiple backends. The release also covers API/Frontend enhancements, security considerations, and provides artifacts (Python wheels, Docker images) along with detailed model support and breaking changes. Overall, it highlights large-scale collaboration and substantial evolution of the vLLM project.