vLLM v0.24.0 Adds MiniMax-M3 Model Support and BF16/FP8 Indexer

Core summary

One sentence to understand this update

vLLM v0.24.0, a significant release with 571 commits from 256 contributors, now supports the new MiniMax-M3 model and includes a BF16/FP8 indexer.

Impact & opportunity

What this could mean

Developers using vLLM can now efficiently deploy and experiment with the MiniMax-M3 model, benefiting from optimized performance through BF16/FP8 indexing, expanding their model choices and capabilities.