Google launches Gemini Decisions API
Google has launched the Gemini Decisions API, which scores a fixed set of options in one forward pass instead of generating text, for cheap, fast routing, policy checks and tool selection.
Google announced the API alongside a new model, Gemini Decision 1 (gemini-decision-1). Both are available through the Gemini API in Google AI Studio.
The model does not decode answers token by token. It scores the candidate options and returns probabilities with a confidence score from 0 to 1, so developers can send unclear cases to a fallback or a human reviewer. Each request takes one multimodal input of up to 256K tokens, which can mix text, images, audio, video and PDFs, plus up to 1,000 independent questions. There are three question types: yes/no checks, multiple choice with 2 to 255 options, and scores on an ordered rubric.
Google says end-to-end latency is 150–250 ms, of which 50–100 ms is model compute, for contexts up to 16K tokens. It says latency stays nearly flat from 1 to 100 questions per request. Pricing is $0.10 per million input tokens with no charge for output tokens, and the shared input is billed once per request. On its own benchmark suite, Google says this makes the API nearly 5x cheaper than Gemini 3.5 Flash-Lite through the Interactions API. It also says the model leads that suite and has the lowest calibration error. The post does not name the benchmarks or the models compared.
Several Google products already use the API. Gemini Notebook uses it to route queries to model tiers, cutting latency on everyday requests by up to 40%. Stitch, CC from Google Labs (for emoji selection) and Google Search (as a safety layer for MCP tools) also use it.
AI Studio offers a free tier for prototyping.