The Beta Summarize Conversation API is no longer available

We are happy to announce that the new Contextual Answer version (v2.1.0) is now available on Amazon SageMaker JumpStart! 🎉

The main focus of Jurassic 2.3.0 includes improvements in grounded generation, making it more reliable and less prone to hallucinations. The new version demonstrates increased performance in output based on a specified context, such as an article, document or report. Popular applications of grounded generation are observed in tasks such as contextual question answering, summarization, and description generation.

We are happy to announce the release of Contextual Answers 2.1.0!

We are pleased to announce the launch of our new ai21-tokenizer client library. This library allows for local tokenization, eliminating the need to rely on the /tokenizer API endpoint. Users can now efficiently perform tokenization directly within their local environment, providing a secure method to determine the number of tokens for a given prompt or answer.

Effortlessly log and organize your playground completions into a data collection.

A new patch release (v2.2.001) of Jurassic-2 Ultra is now live in Amazon SageMaker JumpStart, enabling Jurassic-2 Ultra to run on p4d.24xlarge instances with an expanded context window of 8k tokens (up from the previous 4k context window).

This version of the Contextual Answers model includes the following improvements:

We're thrilled to present our brand new playground, offering you not just one, but multiple playground views. The new playground lets you easily evaluate and engage with our diverse range of foundation and task-specific models, speeding up your prompt engineering and model customization journey.

This new version of Jurassic demonstrates improved capabilities across all tasks and applications, featuring enhanced control over outputs and tighter adherence to specifications. From paying attention to precise instructions, such as including or excluding particular words in the generated text, to ensuring the desired output format, it excels in capturing details like brand voice and product specifics. The model now also exhibits enhanced tolerance to common input imperfections, such as typing and grammatical mistakes.