Google Gemini Decisions API & Decision 1 Model: Pricing
Gemini Decisions API: key facts
| Item | Detail | Source |
|---|---|---|
| Model ID | gemini-decision-1, served through the Gemini Decisions API | Google blog, Nov 11, 2026 |
| Launch date | November 11, 2026, in the Gemini API and Google AI Studio; Google Cloud support listed as coming soon | Google blog, Nov 11, 2026 |
| Question types | Yes/No probability, multiple choice with 2 to 255 options, and ordered rubric scores; every answer carries a 0-1 confidence score | Gemini API documentation |
| Speed and limits | 150-250 ms end to end; up to 256K input tokens and up to 1,000 questions per request | Google blog, Nov 11, 2026 |
| Pricing | $0.10 per 1M input tokens, $0 for output; shared input billed once per request; a no-cost tier for prototyping in Google AI Studio | Google blog, Nov 11, 2026 |
| Integrations | google-genai SDKs plus rollouts with OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode | Google blog, Nov 11, 2026 |
Google Gemini Decisions API and the Gemini Decision 1 Model: How It Works, Pricing and the Stocks Exposed
On November 11, 2026 Google DeepMind launched the Gemini Decisions API, powered by a new model called Gemini Decision 1 (gemini-decision-1). Instead of writing text, it scores a fixed set of answers and returns probabilities in 150-250 ms. Here is what the model does, what it costs, its limits, and which listed companies it affects.
Gemini Decision 1 in short: a model that picks, it does not write
Gemini Decision 1 is a Gemini model built for closed-set decisions. You send it content plus questions with fixed answers (yes/no, multiple choice, or a scored rubric), and it scores every candidate answer in a single forward pass, returning a probability for each option and a confidence score from 0 to 1. It does not generate free text.
Google says end-to-end latency is 150-250 ms, with median model compute of 50-100 ms, and that latency stays almost flat from 1 to 100 questions per request. One request can carry up to 256K tokens of text, images, audio, video or PDFs and up to 1,000 independent questions about that input.
The target jobs are the small, high-volume decisions inside AI products: choosing which tool an agent should call, routing a query to a cheaper or stronger model, policy and safety checks, and grading retrieved context. Low-confidence answers can be sent to a larger model or a human reviewer instead of being trusted blindly.
Gemini Decision 1 vs general chat models: what Google claims
Google says Gemini Decision 1 leads its benchmark suite covering agent routing, reasoning, ranking and multimodal understanding, and has the lowest calibration error of the models compared, which means its stated confidence tracks how often it is actually right. Dataconomy reported a 92.2% score on JevBench against 90.9% for OpenAI's GPT-6 Luna; these are vendor-reported numbers, not independent tests.
The bigger change is cost. Google says the same benchmark suite runs nearly 5x cheaper than Gemini 3.5 Flash-Lite through the Interactions API, because no output, thinking or transcription tokens are billed and a shared document is charged once no matter how many questions are asked about it.
Google already uses the model inside its own products: query routing in Gemini Notebook (up to 40% lower latency on everyday queries), a safety layer in Google Search that flags adversarial behaviour in MCP tools, wireframe editing in Stitch and emoji selection in a Google Labs app.
Listed companies most exposed to the Gemini Decisions API
| Company | Link to the launch | What to watch next |
|---|---|---|
| Alphabet (GOOGL) | Owner of Gemini; a cheap decision layer pulls more agent and app traffic onto the Gemini API and, later, Google Cloud. | Google Cloud revenue and backlog commentary, and the date Decisions arrives in Google Cloud. |
| Broadcom (AVGO) | Co-designs Google's TPU chips; more high-volume inference on Google infrastructure supports TPU demand. | AI semiconductor revenue guidance and comments on Google TPU volumes. |
| Microsoft (MSFT) | Main cloud partner of OpenAI, whose models are the benchmark Google compares against on price and accuracy. | Any OpenAI or Azure response on pricing for routing and classification workloads. |
| Amazon (AMZN) | AWS Bedrock hosts Anthropic and other models used for the same routing and moderation jobs. | Bedrock price changes and new small-model launches from AWS or Anthropic. |
| NVIDIA (NVDA) | Cheaper inference per decision can raise total AI usage, but work running on TPUs is work not running on NVIDIA GPUs. | Inference mix commentary from hyperscalers in the next earnings round. |
Data in practice: tracking the companies behind the AI model race
Price cuts in AI show up first in what companies disclose and what governments buy. Gemral Edge Feed 1 links US federal contract awards, pre-award notices, lobbying filings, congressional trade disclosures and insider filings to the listed companies involved, and every record is tied to the company and ticker that were valid on the day it became public.
When a model like Gemini Decision 1 lowers the cost of agent routing and moderation, the first public traces are cloud and AI-service contract awards and filings from the chip and cloud suppliers that benefit. Edge dates every record to the day it was published and applies that lag before any join, so launch news is never matched to data that came later. Details at /edge/data-coverage and /edge/methodology.
Gemini Decisions API: what to be careful about
The facts here come from Google's November 11, 2026 announcement and developer documentation. Benchmark results are Google's own, the API is in the Gemini Developer API rather than Google Cloud for now, and free-tier limits can change, so check current terms in Google AI Studio before you build on it.
The model only chooses among answers you give it. If the right answer is not in your option list, it will still pick one, so keep a fallback path for low-confidence results. Nothing on this page is investment advice.
Related research on Gemral Edge
| Page | What it covers |
|---|---|
| Google Gemini 4 Carbon model | Google's frontier model released Nov 10, its pricing and the stocks exposed. |
| OpenAI GPT-6.2 release date and pricing | What is confirmed versus rumoured about OpenAI's next model. |
| Anthropic Claude Fable 5.2 and IPO stocks | Anthropic's model line-up and the companies tied to it. |
| Anthropic IPO odds and timeline | S-1 filing, Nasdaq ticker ANT and the oversubscribed roadshow. |
| Broadcom custom ASICs and Google TPUs | The chip supply chain behind Gemini inference. |
| Bittensor TAO and decentralized AI compute | The crypto side of the AI compute market. |
| Google Project Astra and Gemini Live glasses | Google's Gemini hardware push and its suppliers. |
Access Real-Time Terminal Intelligence & Quantitative Signals
Unlock instant Telegram alerts, full congressional portfolio archives, and algorithmic catalyst radar.
Upgrade to Gemral Edge Pro ($39/mo)Frequently asked questions
What is the Google Gemini Decisions API?
It is a Gemini API endpoint, launched November 11, 2026, for closed-set decisions. Powered by the Gemini Decision 1 model, it scores fixed answer options and returns probabilities and confidence scores instead of generated text.
What is Gemini Decision 1?
Gemini Decision 1 (model ID gemini-decision-1) is the model behind the Decisions API. It handles text, images, audio, video and PDFs up to 256K tokens and answers up to 1,000 yes/no, multiple-choice or rubric questions per request in about 150-250 ms.
How much does the Gemini Decisions API cost?
Google lists $0.10 per 1M input tokens and $0 for output tokens, with shared input billed once per request. Google AI Studio also offers a no-cost tier for prototyping.
Is the Gemini Decisions API available on Google Cloud or Vertex AI?
At launch it runs in the Gemini API through Google AI Studio. Google says support in Google Cloud is coming soon but has not given a date.
Which stocks does the Gemini Decisions API affect?
Alphabet (GOOGL) is most directly exposed as the owner of Gemini, with Broadcom (AVGO) linked through TPU chips. Microsoft (MSFT), Amazon (AMZN) and NVIDIA (NVDA) are affected through competition for AI inference spending. This is information, not investment advice.
Risk Disclaimer
Trading and investing in digital assets, financial instruments, and predictive events involve substantial risk of loss and are not suitable for every investor. The predictive intelligence, probability distributions, historical precedents, and scenario modeling presented on this page are compiled for informational and research purposes only and do not constitute financial, investment, legal, or tax advice. Past performance and statistical precedents do not guarantee future outcomes. Always conduct independent due diligence before committing capital.