Make Your Mixture-of-Models Programmable.
Documentation | Playground | Blog | Publications | Hugging Face | Slack
vLLM Semantic Router is a programmable routing layer for building Mixture-of-Models systems across heterogeneous LLM infrastructure. It evaluates request signals, user preferences, and application policies to select—or compose—the right model path for each request.
Use it to improve quality, cost, latency, privacy, and safety without hard-coding routing logic into applications.
| Dimension | Fragmented today | With vLLM SR |
|---|---|---|
| Models | Models specialize in different work. | Compose personalized model paths. |
| Compute | GPUs, accelerators, edge, and cloud coexist. | Route across heterogeneous compute. |
| Location | Inference spans edge, private, and cloud. | Keep data within its boundaries. |
| Preference | "Best" changes by user and workload. | Make every preference executable. |
curl -fsSL https://vllm-sr.ai/install.sh | bash -s -- --channel stableFor pip, uv, or agent-driven installation, see the Installation Guide.
Try the online playground at https://app.vllm-sr.ai/playground.
Credentials:
- Username:
[email protected] - Password:
vllm-sr-read
- [2026/07/21] New Blog: Beyond a Single Model: Building Mixture-of-Models Systems with vLLM Semantic Router
- [2026/06/29] New Blog: Micro-Agent: Beat Frontier Models with Collaboration inside Model API
- [2026/06/16] New Blog: Beyond One Model: Fusion in vLLM Semantic Router
- [2026/06/05] v0.3 Released: vLLM Semantic Router v0.3 Themis: From Signals to Stateful Production Routing
Earlier announcements
- [2026/03/24] Vision Paper Released: The Workload-Router-Pool Architecture for LLM Inference Optimization
- [2026/03/10] v0.2 Released: vLLM Semantic Router v0.2 Athena Release
- [2026/02/27] White Paper Released: Signal Driven Decision Routing for Mixture-of-Modality Models
- [2026/01/05] Iris v0.1 Released: vLLM Semantic Router v0.1 Iris: The First Major Release
- [2025/12/16] Collaboration: AMD × vLLM Semantic Router: Building the System Intelligence Together
- [2025/12/15] New Blog: Token-Level Truth: Real-Time Hallucination Detection for Production LLMs
- [2025/11/19] New Blog: Signal-Decision Driven Architecture: Reshaping Semantic Routing at Scale
- [2025/11/03] Paper Published: Category-Aware Semantic Caching for Heterogeneous LLM Workloads
- [2025/10/27] New Blog: Scaling Semantic Routing with Extensible LoRA
- [2025/10/12] Paper Accepted: When to Reason: Semantic Router for vLLM
- [2025/10/08] Collaboration: vLLM Semantic Router with vLLM Production Stack Team.
- [2025/09/01] Released the project: vLLM Semantic Router: Next Phase in LLM inference.
More announcements are available on the Blog and Publications pages.
For questions, feedback, or to contribute, please join the #semantic-router channel in vLLM Slack.
Track contributors, workgroups, and weekly activity at community.vllm-sr.ai.
We host two monthly community meetings across APAC and the Americas:
- APAC-friendly meeting — second Wednesday of the month: 9:00-10:00 AM Singapore time (UTC+8; the same local time in Beijing)
- Americas-friendly meeting — fourth Wednesday of the month: 8:00-9:00 PM Eastern Time (
America/New_York) / 5:00-6:00 PM Pacific Time
If you want to contribute, start with CONTRIBUTING.md.
For repository-native development workflow and validation commands, use AGENTS.md as the entrypoint and tools/agent/docs/README.md as the canonical index.
If you find Semantic Router helpful in your research or projects, please consider citing it:
@misc{semanticrouter2025,
title={vLLM Semantic Router},
author={vLLM Semantic Router Team},
year={2025},
howpublished={\url{https://github.com/vllm-project/semantic-router}},
}
We are grateful to our sponsors who support us:
AMD provides us with GPU resources and ROCm™ software for training and researching frontier router models, enhancing E2E testing, and building the online models playground.
