GoModel

GoModel is an MIT-licensed, self-hosted AI gateway in Go that fronts 31 LLM providers behind OpenAI and Anthropic compatible endpoints with caching, failover, budgets, virtual keys, and audit logs.

GoModel positions itself as an open-source, self-hosted alternative to hosted routing services. The core is MIT licensed and deploys as a single Go binary, via Docker Compose, or through Helm, and the vendor states it works air-gapped and never phones home, which is the point of difference against a managed gateway. Functionally it consolidates 31 providers, including OpenAI, Anthropic, Gemini, Vertex AI, Bedrock, Azure OpenAI, Groq, Fireworks, DeepSeek, xAI, and local runtimes such as Ollama and vLLM, behind endpoints that are compatible with both the OpenAI and Anthropic API shapes. Around that it layers the operational controls teams actually need: virtual keys, per-key budgets and rate limits, caching, failover and load balancing, audit logs, and usage and cost tracking, with OpenTelemetry and Prometheus for observability. A commercial Pro tier adds prompt compression, OIDC SSO with group gating, per-child quota templates, and intelligent routing in beta, priced at 499 dollars per month or 4,999 dollars per year with unlimited requests, seats, nodes, and environments.