vllm-project/vllm-omni

[Aligned with vllm 0.24.0 now][RFC]: vLLM-Omni 2026 Q2 Roadmap

Closed

#2,136 opened on Mar 24, 2026

View on GitHubContributor guide
 (11 comments) (7 reactions) (1 assignee)Python (1,067 forks)github user discovery
good first issuehelp wantedroadmap

Repository metrics

Stars
 (4,990 stars)
PR merge metrics
 (PR metrics pending)

Description

Motivation.

This issue outlines the development roadmap for vllm-omni in Q2 2026. We are collecting your feedbacks and finalize it later.

Proposed Change.

1) Model Support

  • Omni Models #2207
  • TTS Models #2115
  • Diffusion Models #2226 including world models and vla models

2) Feature Support

  • Quantization #1854
  • Prefix Cache #1184
  • KV offload #1330
  • Model Runner V2 #1770
  • Streaming audio/video input/output #2208
  • RL integration with veRL https://github.com/verl-project/verl/issues/5755
  • Diffusion Optimizations #1217
  • Diffusion continuous batching: #1769
  • Diffusers backend #2403

3) Large Scale Deployment (Disaggregated Serving)

  • EPDG disaggregation: x(E)y(P/D)zG #2336
  • targeting models: Bagel/HY-Image3.0/Qwen-Omni/Ming-Omni

4) Hardware Support

  • ROCm (AMD): #2413
  • Intel XPU: #2574
  • Ascend NPU: #2223
  • MUSA: #2347

5) CI/CD & Usability

  • Quality Gates: L1-L5 level tests enhancement&refactor #2299 #4197 #4254
  • Benchmarking performance: vllm-omni-kanban
  • Developer Usability: vllm-omni-skills
  • User guidance: vllm/recipes

Feedback Period.

No response

CC List.

@ywang96 @Gaohan123 @youkaichao @Isotr0py @ZJY0516 @david6666666 @SamitHuang @linyueqian @wtomin @tzhouam @gcanlin @tjtanaa @xuechendi

Any Other Things.

No response

Before submitting a new issue...

  • Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

Contributor guide

Research direction
Review the proposed roadmap items and provide feedback on priorities or suggest additional features.
Tech stack
python
Domain
backendaimachine learning
Issue type
Research
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Over 1 week
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
PythonGit
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
20