The latest model from the Amazon Titan family, called Amazon Titan Text Premier, is now available on Amazon Bedrock. This advanced and high-performance addition is ideal for text generation business applications, optimizing tasks such as text generation, classification, question answering, and information extraction. Developed with responsible AI practices, Amazon Titan Text Premier excels in offering exceptional generative AI capabilities at scale.
Exclusive to Amazon Bedrock, this model supports a wide range of text tasks and has been optimized for key features, such as retrieval-assisted text generation (RAG) and interactive agent creation. These integrations enable the development of advanced applications, such as AI assistants that can use APIs and custom documents.
The Amazon Titan family offers models with context windows ranging from 4K to 32K, allowing for rich text and code generation, API orchestration, and agent-based applications. The following table summarizes the characteristics of these models:
| Model | Availability | Context Window | Languages | Functionality | Custom Fine-Tuning |
|—————–|:——————:|:——————–:|———|:———————————:|:——————:|
| Amazon Titan Text Lite | GA | 4K | English | Code, enriched text | Yes |
| Amazon Titan Text Express| GA (English) | 8K | Multilingual (100+ languages) | Code, enriched text, API orchestration | Yes |
| Amazon Titan Text Premier| GA | 32K | English | Text generation business applications, RAG, agents| Yes (preview) |
This model is designed for business applications, offering optimized performance and cost-effectiveness with a maximum context length of 32,000 tokens. Amazon Titan Text Premier enables the development of custom agents and supports the creation of interactive AI assistants that can be integrated with APIs and existing data sources. It is available for customization with own datasets, enhancing its performance in specific use cases.
Internal testing has shown that custom-tuned Amazon Titan Text Premier models significantly outperform their baseline in various tasks. To test model customization, it is recommended to contact the AWS account team.
Amazon Titan Text Premier has demonstrated strong performance on various critical public benchmarks for business applications. For example, it has achieved 92.6% on the HellaSwag benchmark for common reasoning, outperforming competitor models. It has also shown outstanding results in reading comprehension and multi-step reasoning. However, customers are encouraged to conduct their own assessments to ensure the model meets their specific requirements.
To start using Amazon Titan Text Premier on Amazon Bedrock, model access must be enabled through the Amazon Bedrock console. The model’s easy accessibility is complemented by the ability to invoke it through REST API or the AWS SDK.
The integration of Amazon Titan Text Premier with Knowledge Bases and Agents for Amazon Bedrock enables the creation of high-quality RAG applications and allows agents to perform complex tasks using specific APIs and knowledge bases. These features represent a significant advancement for AI business applications.
With the general availability of Amazon Titan Text Premier, Amazon hopes that organizations can optimize their workflows and improve their operations and customer experiences through advanced linguistic AI capabilities. For more information on the Amazon Titan model family and pricing details, visit the corresponding Amazon pages.
via: MiMub in Spanish
