Jamba 1.5 household
We are happy to announce that the powerful new Jamba 1.5 family of large language models (LLMs) from AI21 Labs is now available on Amazon Bedrock. These models have made substantial advancements in long-context language capabilities, making them faster, more efficient, and valuable for a wide range of applications. The Jamba 1.5 Mini and Jamba 1.5 Large are the two variations of the Jamba 1.5 family. Document objects, function calls, and a 256K token context window with structured JSON output can all be consumed by any model.AI21 Labs specializes in developing artificial intelligence (AI) systems and foundation models for the workplace. AI21 Labs and AWS are strategically collaborating to empower customers across several sectors to design, build, and expand generative AI solutions that solve urgent problems and inspire innovation. Customers can employ AI21 Labs' state-of-the-art, production-ready models in conjunction with Amazon's specialized services and strong infrastructure to use LLMs in a safe setting to impact people's information processing, communication, and learning going ahead.
Jamba 1.5: What is it?
Jamba 1.5 models use a unique hybrid design that combines the technology of the Structured State Space Model (SSM) with the architecture of the transformer model. This new approach preserves the high-performance characteristics of traditional transformer models while enabling Jamba 1.5 models to handle extended context windows up to 256K tokens. The whitepaper Jamba: A Hybrid Transformer-Mamba Language Model goes into further depth about this hybrid SSM/transformer design.Amazon Bedrock is now compatible with the following two AI21 Jamba 1.5 models:
Jamba 1.5 Large: Large works incredibly well on complicated reasoning tasks across all prompt lengths, making it the ideal option for applications that require high-quality results on both long and short inputs.
Jamba 1.5 Mini: Low-latency processing of lengthy prompts by Jamba 1.5 Mini enables quick evaluation of huge documents and data.
The following are the primary benefits of the Jamba 1.5 models:
- Extended context handling: Jamba 1.5 models can improve the efficiency of enterprise applications, such as agentic and RAG processes and large document summarizing and analysis, thanks to their 256K token context length.
- Multiple languages are supported, including Hebrew, Arabic, German, Dutch, Spanish, French, Portuguese, Italian, and English.
- Kind to developers: It can parse document objects, perform functions, and handle structured JSON output automatically.
- Efficiency and speed: AI21 assessed the Jamba 1.5 models and found that, when it comes to inference on long contexts, the models perform up to 2.5 times better than other models of comparable sizes.
Utilize Jamba 1.5 models immediately on Amazon Bedrock.
Go to the Amazon Bedrock console, choose Model access from the pane on the bottom left, then submit a request to access Jamba 1.5 Small or Jamba 1.5 Large to start using the new Jamba 1.5 models.Bedrock Jamba 1.5 models on Amazon
To try the Jamba 1.5 models, select the Text or Chat playground from the Amazon Bedrock dashboard's left menu pane. Next, select the model by clicking Select model, selecting AI21 as the category, and selecting Jamba 1.5 Mini or Jamba 1.5 Large.You can create your apps in a number of programming languages and access available models using AWS SDKs.
The Jamba 1.5 models are perfect for use cases including long document question-answering, compliance analysis, and paired document analysis. They can quickly process long or complicated papers, analyze information from multiple sources, and assess if a piece complies with specific guidelines. Sample code is available in the GitHub repository AI21-on-AWS. For additional information about prompting Jamba models effectively, see the documentation provided by AI21.
0 Comments