TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Google announced the Gemini Decisions API, powered by Gemini Decision 1, for evaluating up to 1,000 yes/no, multiple-choice and rubric questions against shared multimodal input. The company reports 150–250 ms end-to-end latency and pricing of $0.10 per million input tokens, with no charge for output tokens; independent benchmark results are not provided in the source.
Google announced the Gemini Decisions API on Nov. 11, introducing an endpoint powered by its Gemini Decision 1 model for scoring up to 1,000 yes/no, choice or rubric questions against one shared input. The API accepts text, images, audio, video and PDFs and returns probabilities and confidence scores, giving developers a way to route bounded decisions without generating a separate text answer for each one.
Google says the endpoint can process shared inputs of up to 256,000 tokens and answer up to 1,000 independent questions in one request. Its three question types cover yes/no conditions, selection among 2 to 255 options, and ordered rubric scores. Answers include confidence values from 0 to 1; choice questions also return normalized probabilities for each candidate, while rubric questions return probabilities by level and a weighted score.
The company reports 150–250 milliseconds end-to-end latency, with 50–100 milliseconds attributed to model computation. It says latency remains nearly flat from one to 100 questions per request, and that multimodal inputs and contexts up to 16,000 tokens were included in its performance description. Those figures are Google’s claims; the source does not provide an independent evaluation or full benchmark results.
Pricing is listed as $0.10 per million input tokens and no charge for output tokens. Google says the shared input is billed once per request, regardless of how many questions are evaluated, with each question adding only a small amount to the cost. The API is available through the Gemini API in Google AI Studio, which also has a no-cost tier for prototyping.
Gemini Decisions API Fast, calibrated multimodal decisions
Google’s API scores up to 1,000 defined questions against shared multimodal input. Its speed, pricing and calibration claims await independent evaluation.
How it works
Gemini Decision 1 scores defined answers in one pass, instead of generating a separate text response for each question.
Provide text, images, audio, video or PDFs once for the batch.
Ask yes/no, multiple-choice or ordered rubric questions.
Get answers with confidence values; choice questions also return probabilities.
Where it could help
Several checks on the same material may need fewer repeated model calls.
Choose a path
Route requests to a model tier or fallback when confidence is low.
Check policy
Score defined conditions across documents, images or recordings.
Grade against a rubric
Return ordered scores for context without generating a separate answer for each check.
What the announcement supports
These figures and deployments are reported by Google; the source gives no independent evaluation.
- Latency stays nearly flat from one to 100 questions.
- Gemini Notebook reports up to 40% lower latency for everyday queries.
- Shared input is billed once per request.
- Benchmark names, datasets and detailed results.
- How confidence calibration holds across tasks and customer data.
- Performance across modalities and production workloads.
Try it in context
Developers can prototype through the Gemini API in Google AI Studio’s no-cost tier.
Where Batch Decisions Could Help
The API targets systems that must choose among defined actions quickly: selecting a model tier, checking a policy condition, routing a request or grading context. A single multimodal input can support multiple checks, such as assessing a claim document alongside damage photos or matching a call recording with account information. That could reduce repeated model calls when a workflow needs several answers about the same material.
Confidence scores may help developers send uncertain cases to a fallback model or human reviewer. They do not, by themselves, show that a decision is correct or safe. The practical value will depend on whether the scores are calibrated for a team’s data and whether the speed and cost claims hold under its workloads.
Google says Gemini Notebook uses the endpoint to route queries and reports up to 40% lower latency for everyday queries. It also describes use in Stitch, CC from Google Labs and a safety layer for MCP tools in Search. These are company-reported deployments; the announcement does not include outside customer results or detailed measurements for each use.
Products Worth Considering
Samsung Galaxy Tab S7 11-inch Android Tablet 128GB Wi‑Fi with S Pen Mystic Black
LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz
Amazon Kindle Scribe (3rd Generation), 64GB, 11-inch Display, Graphite, with Premium Pen
As an affiliate, we earn on qualifying purchases.
A Different API Pattern
Many generative model APIs produce a response as a sequence of tokens. Google describes Gemini Decision 1 as scoring candidate options in a single forward pass, without token-by-token answer generation. The endpoint is presented as a purpose-built option when a system already has a finite set of actions or categories to choose from.
Google compares the Decisions API with Gemini 3.5 Flash-Lite used through its Interactions API, saying the new setup was nearly five times cheaper across a shared benchmark suite. The source does not name the benchmarks, publish scores, or explain the comparison’s workload and cost assumptions in detail. The API’s claims should be read as the company’s launch figures until additional methodology or independent evaluations are available.
““Rather than generating answer text token by token, Gemini Decision 1 scores your candidate options in a single forward pass.””
— Google, in its announcement
Benchmark Details Still Missing
The announcement does not publish benchmark names, datasets, scores, confidence intervals or an independent assessment of the claimed low calibration error. It also does not specify how performance varies across each supported modality, beyond its general latency and context statements. The reported cost comparison lacks enough detail to reproduce it from the source.
Google describes confidence scores as ranging from 0 to 1, but the announcement does not explain how calibration was measured or whether those scores remain reliable across different tasks and customer data. It also leaves open how the API handles ambiguous inputs, question limits above 100 in practice, and latency under heavier real-world traffic. Those details matter for teams considering the scores in safety checks or other consequential workflows.
Testing in Real Workflows
Developers can begin prototyping through the no-cost tier in Google AI Studio and access the endpoint through the Gemini API. The next useful evidence will be documentation and evaluations that spell out benchmark methods, calibration results and performance across modalities, as well as customer experience at production scale.
Teams evaluating the service can compare its response times and costs against their existing classifiers or generative calls, then check whether confidence scores reliably identify cases that need review. Google has not announced a date for additional benchmark disclosures or broader product milestones in the source material.
Where I land
I see the API as a potentially useful tool for workflows where many decisions share the same input and the available answers are already defined. If Google’s latency and pricing figures hold for customer workloads, batching could simplify routing and evaluation while reducing repeated calls.
The strongest counterargument is that fast scoring and a confidence value do not establish reliable decisions. Without published benchmark methods and calibration results, teams cannot tell from the announcement how well the model handles unusual cases or whether its probabilities support safe escalation decisions. I would raise my assessment if independent evaluations showed dependable calibration across modalities and real production workloads; evidence of uneven results would make me more cautious.
Key Questions
What is the Gemini Decisions API?
It is a Google API endpoint powered by Gemini Decision 1 that scores bounded decisions, such as yes/no checks, multiple-choice selections and ordered rubric ratings, against shared input.
What kinds of input does it support?
Google says it accepts text, images, audio, video and PDFs in a shared input of up to 256,000 tokens.
How much does it cost?
The announced price is $0.10 per million input tokens and $0 for output tokens. Google says the shared input is billed once per request; the source does not give a complete per-question cost breakdown.
Are the performance claims independently verified?
The source provides Google’s reported latency, calibration and cost claims, but no independent evaluation or detailed benchmark methodology.
Source: Google AI
