LIVE
Nvidia CEO Jensen Huang Steps Down, CFO Colette Kress Appointed Interim Chief Nvidia Chief Executive Jensen Huang Faces Sudden Scrutiny Amid Market Anxieties South Korean AI and Deep Tech Startups Target US Market Expansion Palantir CEO Alex Karp Defends Defense Tech Firm as 'Progressive' and Deterrent to War Flight Attendants Secure Privacy Safeguards in Google Data Purchase from Bankrupt Spirit Airlines IBM Unveils Quantum Nighthawk r3 Processor Aimed at 10,000 Operations UK Intelligence Services Face Scrutiny Over Artificial Intelligence Oversight Darling Project Adopts AI Contributions Policy and Adds ARM64 Support Nvidia Faces Uncertainty as CEO Jensen Huang's Future Hangs in the Balance Mathematician Uses Coding Agents to Revamp Interactive Logic Textbook into 'Lean Grimoire' Nvidia CEO Jensen Huang Steps Down, CFO Colette Kress Appointed Interim Chief Nvidia Chief Executive Jensen Huang Faces Sudden Scrutiny Amid Market Anxieties South Korean AI and Deep Tech Startups Target US Market Expansion Palantir CEO Alex Karp Defends Defense Tech Firm as 'Progressive' and Deterrent to War Flight Attendants Secure Privacy Safeguards in Google Data Purchase from Bankrupt Spirit Airlines IBM Unveils Quantum Nighthawk r3 Processor Aimed at 10,000 Operations UK Intelligence Services Face Scrutiny Over Artificial Intelligence Oversight Darling Project Adopts AI Contributions Policy and Adds ARM64 Support Nvidia Faces Uncertainty as CEO Jensen Huang's Future Hangs in the Balance Mathematician Uses Coding Agents to Revamp Interactive Logic Textbook into 'Lean Grimoire'

Google Launches Gemini Decisions API and Decision 1 Model for Fast Multimodal Evaluation

🤖 AI Models 1 min read 👁 0

Google has unveiled the Gemini Decisions API powered by the new Gemini Decision 1 model, designed to execute rapid, closed-set evaluations across multiple data modalities.

Google Launches Gemini Decisions API and Decision 1 Model for Fast Multimodal Evaluation🤖AI Models

Photo: Hacker News Front Page

Key points

  • Gemini Decision 1 evaluates up to 1,000 yes/no, choice, and rubric questions in a single pass.
  • Supports multimodal inputs up to 256K tokens, including text, images, audio, video, and PDFs.
  • Priced at $0.10 per 1 million input tokens with zero output token costs.

Google has rolled out the Gemini Decisions API alongside its specialized Gemini Decision 1 model, providing developers with a dedicated endpoint for handling bounded choices, policy checks, and content grading. Accessible through Google AI Studio and the Gemini API, the new tool is engineered to process up to 1,000 questions in a single forward pass.

Rather than relying on traditional token-by-token text generation, the underlying model computes probabilities and confidence scores directly. It supports up to 256K tokens of multimodal input—combining text, images, audio, video, and PDFs—while delivering median end-to-end latencies between 150 and 250 milliseconds. The system evaluates yes-or-no queries, multiple choice options, and rubric scores simultaneously while returning confidence metrics ranging from 0 to 1 for downstream routing.

Pricing for the Gemini Decision 1 model is set at $0.10 per million input tokens with no charges for output tokens, alongside a free prototyping tier. Ecosystem integrations are expanding across platforms including LangChain, Vercel, LiteLLM, and OpenRouter, with upcoming support planned for Google Cloud.

Source Hacker News Front Page — blog.google/innovation-and-ai/technology/developers-tools/gemini-decisi… ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.