Gemini Decisions API Answers 1,000 Questions

Google on Wednesday, Nov. 11, launched the Gemini Decisions API, a new endpoint that answers up to 1,000 yes/no, multiple-choice and scoring questions about a single input in one pass instead of writing text. The service runs on a new model, Gemini Decision 1, and bills only input tokens at $0.10 per million, according to a Google blog post by Google DeepMind product manager Min Choi and principal scientist Felix Yu.

Key takeaways

• One request can pair a text, image, audio, video or PDF input of up to 256,000 tokens with up to 1,000 independent questions.
• Input costs $0.10 per million tokens and output tokens are free; a no-cost tier is available in Google AI Studio for prototyping.
• Google cites end-to-end latency of 150 to 250 milliseconds and says the setup is nearly five times cheaper than Gemini 3.5 Flash-Lite on its benchmark suite.

How the Gemini Decisions API works

The model, called gemini-decision-1 in code, does not generate answers token by token. It scores the candidate options a developer supplies in a single forward pass and returns probabilities, along with a confidence score from 0 to 1 for every answer.

Questions come in three types. A "noul" question estimates the probability that a yes/no condition holds. A "choice" question picks among 2 to 255 options and returns normalized probabilities for each. A "score" question grades input on an ordered rubric that starts at 0 and returns a weighted score.

Inputs can mix modalities, such as a call recording with account data or a scanned claim form with damage photos. Google says the confidence figures let applications send ambiguous cases to a fallback model or a human reviewer.

Gemini Decision 1 pricing and speed

Google prices the model at $0.10 per million input tokens and $0 for output. The shared input is billed once per request regardless of how many questions are asked; each question adds only its own instruction and criteria tokens. OpenRouter's listing for the model shows the same rates.

According to Google, median model compute takes 50 to 100 milliseconds for inputs up to 16,000 tokens, and latency stays nearly flat from 1 to 100 questions per request. The company also says the model posted the lowest calibration error across its benchmark suite. These figures are Google's own and have not been independently tested.

Where Google already uses it

Google says several of its products run on the endpoint. Gemini Notebook uses it to route queries to the right model tier, serving everyday questions with up to 40% lower latency. Google Labs' CC picks emoji reactions from more than 200 options, the Stitch design tool uses it for a rapid wireframing experiment, and Google Search uses it as a safety layer for MCP tools.

Integrations are rolling out through OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode. Support in Gemini in Google Cloud is listed as "coming soon."

Context: OpenAI moved first

The launch follows OpenAI, which made its own Decisions API generally available on Oct. 22. That endpoint runs only on GPT-6 Luna, accepts up to 400 questions per request and also charges $0.10 per million uncached input tokens with free output. Google's version raises the per-request ceiling to 1,000 questions and adds audio, video and PDF input.

Both target the small, frequent calls inside AI agents, such as choosing a tool, checking a policy or deciding whether to escalate. Startups are chasing the same niche; AI Newsway earlier reported that TypeSafe AI left stealth with $40 million on a similar premise.

What's next

The API is available now through the Gemini API in Google AI Studio, with Python, JavaScript and REST guides. The open questions are whether independent tests back Google's speed and calibration claims, and when the endpoint reaches Google Cloud customers.

FAQ

What is the Gemini Decisions API?

It is a Gemini API endpoint that returns probabilities, choices and rubric scores for up to 1,000 questions about one input, rather than generated text.

How much does Gemini Decision 1 cost?

$0.10 per million input tokens, with no charge for output tokens. A free tier is available in Google AI Studio for prototyping.

How is it different from OpenAI's Decisions API?

Both bill input only at $0.10 per million tokens. Google's allows 1,000 questions per request versus OpenAI's 400 and accepts audio, video and PDF input.

References

• Gemini Decisions API: Fast, calibrated multimodal decisions — Google (Nov. 11, 2026)
• Gemini Decision 1 - API Pricing & Providers — OpenRouter (Nov. 11, 2026)
• Everything That Happened in AI Today Thursday, November 12 — The Neuron (Nov. 12, 2026)