v0.17.1

vllm-project/vllmv0.17.1Mar 11, 2026by khluu

AI Summary

A patch release addressing issues with TRTLLM fused MoE, DeepSeek V3.2, and adding support for the Nemotron 3 Super model.

Key Highlights

Fix passing activation_type to trtllm fused MoE
Re-enable EP for trtllm MoE FP8 backend
[Mamba][Qwen3.5] Zero freed SSM cache blocks on GPU
Fix TRTLLM Block FP8 MoE Monolithic
Nemotron 3 Super model support

New Features

Nemotron 3 Super model support
Fixes for nongated fused moe triton

Full Release Notes

This is a patch release on top of `v0.17.0` to address a few issues:
- New Model: Nemotron 3 Super
- Fix passing of activation_type to trtllm fused MoE NVFP4 and FP8 (#36017)
- Fix/resupport nongated fused moe triton (#36412)
- Re-enable EP for trtllm MoE FP8 backend (#36494)
- [Mamba][Qwen3.5] Zero freed SSM cache blocks on GPU (#35219)
- Fix TRTLLM Block FP8 MoE Monolithic (#36296)
- [DSV3.2][MTP] Optimize Indexer MTP handling (#36723)