One endpoint.
Open models.
OpenAI-compatible inference for open-source models.
Point any tool at api.inference.idyl.dev and start building.
Drop-in compatible with the tools you already use
What is idyl.inference?
Why idyl
Built for how
developers actually work.
You already have a preferred IDE, coding agent, and workflow. idyl fits into all of them — no SDK lock-in, no migration.
Drop-in compatible
Same OpenAI API, same SDKs. Change one URL and your existing tools, scripts, and agents work instantly. Zero code changes.
Open-source models
Llama, Qwen, DeepSeek, Mistral, Gemma, Phi — 50+ models from every major open-source family. No vendor lock-in, ever.
Powered by idle compute
Inference runs on the world's underutilized GPUs. Distributed by design — lower cost, higher resilience, no single point of failure.
Full OpenAI API compatibility
Streaming responses, function calling, JSON mode, system prompts — everything you expect from the OpenAI API. If your tool speaks OpenAI, it speaks idyl.
// Streaming + function calling — just works const stream = await client.chat.completions.create( model: "deepseek-v3.2", messages: [ role: "user", content: "..." ], stream: true, tools: [ type: "function", ... ], );
Models
From 8B to 600B+.
Open-source only.
The most capable open models from every major lab — reasoning, code, chat, vision. New models land within days of release.
Getting Started
Three steps. That's it.
Get an API key
Sign up free. No credit card, no commitment. Your key is ready in seconds.
Point your tools
Set your base URL to api.inference.idyl.dev. Same OpenAI SDK. Same everything.
Start building
Claude Code, OpenCode, OpenClaw, cURL, Python — whatever you use, it just works.