Classify 6700 pages for $1 with a Jev-compatible VLM API (twitter.com)

A new API from @typesafeai has been launched that allows developers to classify documents using open-weight Vision Language Models (VLMs) at an unprecedented low cost. For just $1, users can classify over 6,700 pages, while an extensive 1,651-page run of the OmniDocBench only costs 25 cents. The API operates by reading pages as pixels, eliminating the need for complex OCR to LLM conversions, and delivers a calibrated decision for each page in roughly 180 milliseconds. This development is significant for the AI/ML community as it lowers the barriers to entry for document classification workflows, encouraging developers to reconsider data processing strategies with a more cost-effective approach. By positioning vision as a central feature in API design, it enables more seamless integration into various applications while simplifying the request and response structure. Overall, this advancement streamlines tasks for organizations working with large volumes of documents, fostering innovation and efficiency in automation and AI-driven solutions.