# vLLM v0.8.3 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.8.3
- **Released:** Apr 06, 2025
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.8.3/

## Executive Summary (TL;DR)
vLLM v0.8.3. Source highlight: We are excited to announce Day 0 Support for Llama 4 Scout and Maverick (#16104). Please see our blog for detailed user guide.

## New Features
- Add option to use DeepGemm contiguous grouped gemm kernel for fused MoE operations. (#13932)
- Add Reasoning Parser for Granite Models (#14202)
- Add Phi-4-mini function calling support (#14886)
- Add custom allreduce support for ROCM (#14125)
- Add minimum version for `huggingface_hub` to enable Xet downloads (#15873)
- TPU CI: Add basic perf regression test (#15414)

## Improvements
- Expert parallelism enhancements
- LoRA Scheduler optimization bridging V1 and V0 performance (#15422).
- Improve Memory Usage Estimation (#15671)
- Optimize the all-reduce performance (#15903)
- [Misc] Enhance warning information to user-defined chat template
- [Misc] improve example script output

## Bug Fixes
- Fix CUDA kernel index data type in vllm/csrc/quantization/gptq_marlin/awq_marlin_repack.cu +10
- [Hardware][TPU][Bugfix] Fix v1 mp profiler
- [Kernel] Fix conflicting macro names for gguf kernels
- [bugfix] fix inductor cache on max_position_embeddings
- [Bugfix] Support triton==3.3.0+git95326d9f for RTX 5090 (Unsloth + vLLM compatibility)
- [bugfix] add supports_v1 platform interface

## Upgrade Command
```bash
python -m pip install 'vllm==0.8.3'
```