Google Gemini

Gemini's advantage in a pipeline is context and unit cost. Flash-Lite at $0.10 per million input tokens is the cheapest way to classify, tag or triage a large batch of anything, and the context windows are large enough to put an entire document in a single prompt rather than chunking it. Use it where volume is the problem: labelling every asset a pipeline produces, filtering a feed before an expensive model sees it, or summarising long source material in one pass. Sushibelt talks to Gemini through its OpenAI-compatible endpoint, so it behaves like every other LLM Worker on the canvas. Charged on reported tokens; note that Gemini prices audio input differently from text, and this node sends text.