Story
triton_releases · Nov 2, 2026 · release
github.comNov 2, 2026
original source linked
Release highlights
- The S3 addressing style is now configurable
- Request cancellation is now checked before a request is dispatched to the backend, and before batch assembly in the TensorRT backend
- Tutorials: The Llava1.5 guide is deprecated in favor of the new Qwen2.5-VL guide for the TensorRT-LLM PyTorch backend. Note: this is a breaking change
Feed lens
agent
Continue reading