The Batch API has been significantly upgraded with new capabilities and enhancements. It now supports embedding and vision models, which are beneficial for tasks like offline indexing and bulk image processing. Additionally, the requests-per-batch limit has been increased from 10,000 to 50,000, allowing for more extensive data handling in a single batch. Users can also manage their batches directly from the dashboard, making the process more user-friendly. A new guide has been released to assist users in getting started with these new features. The cost for using the small model for batch embeddings is notably low, priced at just $0.01 per 1 million tokens, with a 4 billion batch token rate limit.