Qwen3.5

Flide is a new inference engine that optimizes decision models by separating the logical tree structure of requests from their physical execution stream. It significantly outperforms existing frameworks in speed and numerical precision for LLM agent systems.

Ovis-Embedding presents a state-of-the-art approach to universal embeddings by integrating text, image, video, and audio into a single shared representation space. The model achieves superior performance in any-to-any retrieval tasks through advanced training strategies like focal loss and embedding distillation.

This technical update outlines a series of low-level optimizations for production-grade AI inference serving. It covers significant advancements in speculative decoding, MoE model support, and GPU kernel efficiency.