Google has launched Jev: its Gemini Decision 1 for rapid evaluation of decisions at a price of 10 cents per million tokens.
This is classified. An analogue of Jev, and an analogue of solutions from Cloudflare, OpenAI and so on. But no one writes that, contrary to the above, it can process video. Here is a video analysis of the video.
And Google's models, starting with Flash, are very good at understanding pure video without parsing subtitles or audio tracks.
The model received native multimodality and a context window of 256,000 tokens. Text arrays, high-resolution images, audio recordings, video streams, and multi-page PDF documents are supported as input data. In Google testing, the architecture outperformed TypeSafe's Jev 1 in both accuracy in classifying complex edge cases and resistance to false positives in moderate and compliant tasks.
Pricing at $0.10 per million tokens with no hidden fees or multimodal charges is the industry's most cost-effective streaming content moderation solution.
Access to the model is already open in Google AI Studio and through the Gemini public API.
In summary:
Batch processing up to 1,000 questions at a time: parallel calibrated evaluation of solutions in a single query.
256K multimodal token window: Native analysis of text, images, audio, video and multi-page PDF files.
$0.10 rate for 1 million input tokens: a tenfold cost reduction compared to universal front-end models.
Get started right away: Google AI Studio integration, ready-made web demos, and public documentation on the Gemini API.