ブログ
プロダクトの更新情報、ガイド、開発の裏側をお届けします。
How to call the Fusion API (OpenRouter)
A practical guide to calling the OpenRouter Fusion API: the endpoint, request shape, custom panels, the judge model, and streaming — with copy-paste examples.
FusionAPI
Fusion API Python example (OpenRouter)
Call the OpenRouter Fusion API from Python using the openai SDK: a minimal example, streaming, and a custom panel via extra_body — copy-paste ready.
FusionAPI
Fusion API Node.js example (OpenRouter)
Call the OpenRouter Fusion API from Node.js / TypeScript: the openai SDK, streaming, a custom panel, and a raw fetch example — copy-paste ready.
FusionAPI
Using the Fusion API in Cursor, Cline, LangChain & Vercel AI SDK
How to plug the OpenRouter Fusion API into Cursor, Cline, LangChain and the Vercel AI SDK — since it is OpenAI-compatible, it drops into each in a few lines.
FusionAPI
Why the Fusion API costs more than a single model
A Fusion API call is billed as the sum of every model in the panel plus the judge — here is how that adds up, the typical multiplier, and when it is worth paying.
FusionAPI
Why running the same model twice makes it smarter
Most of the gain from multi-model Fusion comes from the synthesis step — not model diversity. Even pairing a model with a copy of itself improves results.
FusionAPI
DRACO, explained: how to benchmark deep research
DRACO grades models on 100 deep-research tasks across 10 domains — accuracy, completeness, citations, and penalties for risky claims. Here is how it works.
FusionAPI