# GoModel

> Open-source AI gateway and AI control plane written in Go. A single ~16 MB binary that routes 22 providers (OpenAI, Anthropic, Google Gemini, Vertex AI, Azure OpenAI, Amazon Bedrock, OpenRouter, Groq, xAI, DeepSeek, Ollama, vLLM, and more) behind OpenAI- and Anthropic-compatible APIs.

GoModel unifies access to 700+ models through one endpoint: /v1/chat/completions, the /v1/responses lifecycle, and the Anthropic-compatible /v1/messages, plus embeddings, files, batches, audio, and realtime speech. On top of routing it adds model aliases and virtual models with load balancing, scoped workflows, guardrails, exact-match and semantic response caching, automatic failover with retries and a circuit breaker, budgets and rate limits, per-request usage and cost tracking, audit logs, virtual API keys, an MCP gateway, Prometheus metrics, and an embedded admin dashboard. Storage starts on SQLite and scales to PostgreSQL or MongoDB. In a reproducible June 2026 benchmark, GoModel showed 1.8 ms p50 latency overhead, 4,900 req/s sustained throughput, 37 MB peak RAM, and a 0.56 s cold start.

## Docs
- [Documentation llms.txt index](https://gomodel.enterpilot.io/docs/llms.txt): Auto-updated index of every documentation page in llms.txt format
- [GitHub Repository](https://github.com/ENTERPILOT/GOModel): Source code, issues, releases
- [Documentation](https://gomodel.enterpilot.io/docs): Setup guides and API reference
- [Releases](https://github.com/ENTERPILOT/GOModel/releases): Download binaries and Docker images
- [AI Gateway Benchmark](https://enterpilot.io/blog/benchmarking-ai-gateways-gomodel-litellm-portkey-bifrost-june-2026/): GoModel vs LiteLLM vs Portkey vs Bifrost

## Optional
- [Discord Community](https://discord.gg/gaEB9BQSPH): Community chat and support
- [Issues](https://github.com/ENTERPILOT/GOModel/issues): Bug reports and feature requests