Google Launches Gemini Decisions API at $0.10 per 1M Tokens

Google opened a new Gemini Decisions API on Wednesday, pricing scoring at $0.10 per million input tokens with zero output charges and letting a single request carry up to 1,000 bounded questions over multimodal inputs.

The announcement, posted by Min Choi, product manager at Google DeepMind, and Felix Yu, principal scientist at Google DeepMind, pitches the system as "purpose-built primitives for bounded choices like tool selection, policy checks, model routing, and context grading." The underlying model, Gemini Decision 1, "scores your candidate options in a single forward pass, returning probabilities and confidence scores in 150-250 ms end-to-end latency," with roughly 50-100 ms of that being model compute.

Inputs span text, images, audio, video, and PDFs up to 256K tokens, with per-question contexts capped at 16K. Multiple-choice tasks select "from 2 to 255 choices, returning the winner and normalized probabilities for every candidate," and an ordered-rubric option returns a weighted score.

On cost, Google says the setup is "nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API." The post publishes no accuracy figures against that suite and does not describe what tasks it covers. It landed the same day our tracker logged the MMDM decision-model benchmark, where multi-select accuracy across 28 models topped out at 37%.

Ecosystem integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode alongside the google-genai SDKs; Google Cloud support is listed as coming soon. Decision 1 is already running inside Google on dynamic model routing in Gemini Notebook, the new rapid canvas in Stitch, and Google Search, which "uses the Decision API to detect adversarial behavior" on MCP tools.