The release enables developers in Google AI Studio to process multimodal contexts containing up to 256K tokens across text, audio, video, images, and PDFs. It evaluates up to 1,000 independent questions within a single request using three distinct formats, including multiple-choice selections of up to 255 candidates. At an operational cost of $0.10 per million input tokens with zero output fees, it runs nearly five times cheaper than Gemini 3.5 Flash-Lite on the Interactions API. Internal adoption is already underway, with Gemini Notebook utilizing the endpoint to cut everyday task latency by up to 40%.
Key points
Evaluating options in a single forward pass eliminates output token billing and delivers…: Evaluating options in a single forward pass eliminates output token billing and delivers major cost savings over standard text generation.
Confidence scores returned alongside choices allow automated pipelines to seamlessly trig…: Confidence scores returned alongside choices allow automated pipelines to seamlessly trigger fallbacks or human intervention for edge cases.
Integrating the decision model as a real-time intent router can slash application latency…: Integrating the decision model as a real-time intent router can slash application latency by up to 40% for routine user tasks.
Today, we’re introducing the Gemini Decisions API, a new endpoint for fast, closed-set decisions powered by our new Gemini Decision 1 model (gemini-decision-1), available through the Gemini API in Google AI Studio.
The Decisions API gives you purpose-built primitives for bounded choices like tool selection, policy checks, model routing, and context grading. You can pass one multimodal input across text, images, audio, video, and PDFs (up to 256K tokens) along with up to 1,000 independent questions. Rather than generating answer text token by token, 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 at $0.10 per 1M input tokens with zero output tokens billed.
Strong performance and calibration
On benchmarks spanning agent routing, reasoning, ranking, and multimodal understanding, Gemini Decision 1 leads across the suite, with the lowest calibration error.
Gemini Decision 1 scores options without token-by-token decoding, keeping median latency 50-100 ms of model compute (150-250 ms end-to-end), across multimodal inputs and contexts up to 16K tokens, remaining nearly…
Industry stakeholders are evaluating the operational characteristics and market positioning of new technical assets. Understanding the specific balance of capabilities, execution speed, and associated costs remains vital for determining deployment feasibility. These formal assessments provide organizations with essential criteria to benchmark performance before committing resources.
The enterprise software giant is directing over $1 billion into its Bay Area operations through early 2028 to accelerate artificial intelligence development. As client demand grows among prominent labs like OpenAI and Anthropic, this central research hub solidifies the firm's presence in the city while fueling local commercial real estate.
Engineering teams can now bypass redundant development cycles by leveraging 23 production-ready, open-source building blocks released by AWS Frontier AI Engineering & Services. These vetted frameworks eliminate the security and scaling risks common to custom-built generative solutions. As a result, enterprises are cutting technical implementation schedules from months down to weeks while supporting heavy operational scale.