llm-d
Kubernetes-native distributed inference on top of vLLM
Verdict
Splits prefill and decode into separate deployments behind an inference-aware router. Worth a look once you run many GPUs. Below that scale, continuous batching on plain vLLM is simpler and usually enough.
Other Model Serving
- vLLMProduction
High-throughput LLM inference server with PagedAttention
- OllamaStable
Run open-weight LLMs locally — one-command setup
- LiteLLMProduction
100+ LLM providers behind a single OpenAI-compatible API
- NVIDIA TritonProduction
NVIDIA's production model serving platform — any framework, any GPU
- BentoMLStable
Build, ship, and scale AI applications with Python
- Ray ServeStable
Scalable model serving on Ray distributed compute