Google Launches Gemini Decisions API With Gemini Decision 1 at $0.10/M Tokens

What it does

Google introduced the Gemini Decisions API, using its new Gemini Decision 1 model. This API processes up to 1,000 yes/no, choice, and rubric-based questions per call across multiple input types: text, images, audio, video, and PDFs. It returns calibrated probabilities rather than simple binary outputs, helping users understand confidence levels in model decisions. The pricing is set at $0.10 per 1 million input tokens, making it a cost-efficient option for decision-making applications.

Why it matters

The API shifts the way businesses and developers handle decision-making tasks at scale. By providing calibrated probabilities instead of just answers, it supports more nuanced risk assessment and downstream logic. This can improve automation in areas like content moderation, quality control, or customer support where understanding confidence is vital. The broad input modality support also means it can integrate into diverse workflows involving multimodal data, reducing the need for separate solutions. Finally, the low price point pushes down costs for large-scale decision APIs, increasing accessibility for startups and enterprises alike.

Who it is for

Developers building systems that require high-throughput, probabilistic decision support will find this API especially useful. Automation teams focused on content review, compliance, or complex rule evaluation can leverage the calibrated probabilities to reduce false positives and negatives more effectively. Organizations with multimodal workloads—those working with images, audio, or video alongside text—gain a unified API instead of juggling specialized models. The pricing also makes it feasible for smaller companies to adopt decision AI without heavy upfront investment.

The catch

Calibrated probabilities add complexity to downstream logic, meaning teams need to implement additional evaluation layers rather than treating API responses as binary truths. Handling multimodal data may require pre-processing pipelines, which can complicate integration. At $0.10 per million tokens, the price is competitive, but costs could add up fast in very high-volume scenarios. Users will also need to verify that Gemini Decision 1’s calibration matches their domain requirements, as out-of-the-box models may need fine-tuning for niche use cases.

What to watch next

Attention will fall on how Google develops the Gemini Decisions suite and its integration with the broader Gemini model family. Watch for updates on model improvements, domain-specific tuning, and support for additional question types or input formats. Competitors are likely to respond by pushing more capable decision APIs or lowering prices, shaping the economics of AI-driven automation. Finally, user feedback on calibration reliability and multimodal handling will determine market adoption speed and developer trust in this new offering.

AI Quick Briefs Editorial Desk