Routerly vs SGLang

Routerly is a self-hosted gateway for directing application requests to AI providers, tracking usage costs, and applying budgets. Its routing engine can use up to nine configurable policies, covering model choice, cost, provider health and performance, capability and context, budgets, rate limits, and fairness. It supports OpenAI and Anthropic wire formats, including streaming, so compatible clients can connect by changing the base URL. Listed providers include OpenAI, Anthropic, Google Gemini, Ollama, Mistral, Cohere, xAI, and custom HTTP endpoints. Per-token costs can use built-in pricing presets or custom model prices. Budgets can measure cost, calls, or token usage globally, per project, or per token, over rolling or calendar windows. Projects can have separate bearer tokens, model pools, routing configurations, and budget limits. Routerly stores configuration and usage records in JSON files without requiring an external database. It can be deployed with Docker or an installer. The self-hosted plan costs nothing, carries the AGPL-3.0 license, and adds no markup to API calls.

SGLang is a free, open-source inference framework for serving large language, vision-language and diffusion models. It targets agentic workloads, reinforcement-learning rollouts and large-scale serving, from one GPU to distributed clusters. Runtime features include RadixAttention, prefix caching and multi-GPU parallelism. Listed optimizations include disaggregated prefill and decode, speculative decoding, a zero-overhead scheduler and optimized GPU kernels. Supported models include DeepSeek, Qwen, GPT-OSS, Llama, Mistral and GLM. SGLang works with Hugging Face and OpenAI APIs and exposes OpenAI-compatible endpoints. It installs with pip or Docker and supports self-hosted deployment. Hardware support spans NVIDIA and AMD GPUs, Google TPUs, Intel GPUs and CPUs, Apple Silicon, Huawei Ascend NPUs and Moore Threads GPUs. SGLang Diffusion adds image and video generation. The Apache-2.0 license permits open-source use.