April 21, 2026
Gateway wars go brrr
Show HN: GoModel – an open-source AI gateway in Go; 44x lighter than LiteLLM
Go’s 44x lighter AI switchboard has devs cheering while skeptics poke holes
TLDR: GoModel, a Go-based AI “switchboard” claiming 44x less bloat than LiteLLM, offers one OpenAI-style API for many providers. Comments split between Go fans shading Python, demands for cost tracking and subscription support, and tough questions about semantic caching and comparisons to rival Bifrost—with everyone asking for proof before the hype sticks.
GoModel just crashed the party with a bold claim: an OpenAI-style “one plug to rule them all” AI switchboard written in Go that’s “44x lighter than LiteLLM.” Translation: developers can swap between OpenAI, Anthropic, Google Gemini, xAI (Grok), Groq, Azure, Oracle, Ollama, and more without rewriting their apps. Cue the community fireworks.
The loudest cheer came from the compiled-language crowd: one commenter basically winked, saying it’s “expectable” a Go app would be leaner than Python, tossing a playful jab at LiteLLM. But the honeymoon didn’t last. Another voice demanded: how does this actually stack up against Bifrost, the other Go router? The vibe: if GoModel wants the crown, show the receipts.
Then came the nerdy knife fight: “semantic caching.” A sharp-eyed commenter asked how GoModel avoids serving old answers when the underlying model changes—a classic “cache invalidation is hard” moment that had everyone nodding. Power users dropped feature wishlists too: built‑in cost tracking per model/route (hello, AI bill anxiety), and even support for consumer subscriptions like ChatGPT and GitHub Copilot.
So yes, there’s hype—faster, slimmer, one API to many AIs—but the thread crackled with real‑world demands. The crowd wants benchmarks, receipts, and guardrails before crowning a new gateway king.
Key Points
- •GoModel is an open-source AI gateway written in Go that exposes an OpenAI-compatible API across multiple providers.
- •Deployment is demonstrated with Docker, supplying at least one provider credential; using an env file is recommended for secret management.
- •The gateway auto-detects available providers based on supplied credentials and provides a sample /v1/chat/completions call.
- •A feature matrix details per-provider support for chat, responses, embeddings, files, batches, and passthrough; OpenAI and Azure OpenAI show full coverage.
- •Alternative setups include running from source (Go 1.26.2+), Docker Compose with infra/full-stack profiles, and building a local Docker image; key endpoints for chat, responses, embeddings, and file operations are listed.