Gemini Decision 1: Google's new classifier for content analysis
Google has introduced Gemini Decision 1, a specialized classifier for rapid decision assessment, available at a price of 10 cents per million tokens. The model is designed for streaming content audits and surpasses competitors in accuracy and resistance to false positives.
Multimodality and Video Analysis
Unlike alternatives from Cloudflare or OpenAI, Gemini Decision 1 natively supports video stream processing without the need for prior analysis of subtitles or audio tracks. The model architecture demonstrates high efficiency in understanding raw video, making it a unique tool for multimedia content moderation.
Technical Specifications and Capabilities
The model features a context window of 256,000 tokens and supports the following input formats:
- text arrays
- high-resolution images
- audio recordings
- video streams
- multi-page PDF documents
In moderation and compliance test tasks, the architecture outperformed TypeSafe's Jev 1 model in both the accuracy of classifying complex edge cases and resilience against false positives.
Pricing and Availability
Usage costs $0.10 per million input tokens with no hidden fees or surcharges for multimodality. This solution provides a tenfold reduction in costs compared to universal frontier models, making it the most affordable industrial option for audit tasks.
Key API advantages include batch processing of up to 1,000 questions per request, enabling parallel calibrated assessment of decisions. Integration is available via Google AI Studio and the public Gemini API, where developers can find ready-made web demos and documentation.









