Bedrock
Bedrock
Bedrock is a product by AWS that accelerates development of generative AI applications using FMs through an API, without managing infrastructure. Currently, Bedrock provides access to foundation model from Anthropic, AI21, AWS Titan family and Stability AI. Bedrock also enables us to privately customize FMs using our organization's data.
In the hackathon, we chose to expose those foundation models through Jarvis.
| Model Name | Max Tokens | Model Id | Notes | Use Cases | Classification |
|---|---|---|---|---|---|
| Claude v2.1 | 20k | anthropic.claude-v2.1 | An update to Claude 2 that features double the context window, plus improvements across reliability, hallucination rates, and evidence-based accuracy in long document and RAG contexts. | Question answering, information extraction, removing PII, content generation, multiple choice classification, Roleplay, comparing text, summarization, document Q&A with citation | Text generation, Conversation, Complex reasoning & analysis |
| Claude v2 | 20k | anthropic.claude-v2 | Anthropic's most powerful model, which excels at a wide range of tasks from sophisticated dialogue and creative content generation to detailed instruction following. | Same as Claude v2.1 | Text generation, Conversation, Complex reasoning & analysis |
| Claude Instant v1 | 20k | anthropic.claude-instant-v1 | A faster and cheaper yet still very capable model, which can handle a range of tasks including casual dialogue, text analysis, summarization, and document question-answering. | Same as Claude v2.1 | Text generation, Conversational |
| Jurassic-2 Ultra | 8k | ai21.j2-ultra-v1 | Jurassic-2 Ultra is AI21’s most powerful model offering exceptional quality. Apply Jurassic-2 Ultra to complex tasks that require advanced text generation and comprehension. Popular use cases include question answering, summarization, long-form copy generation, advanced information extraction, and more. | Open book question answering, summarization, draft generation, information extraction, ideation | Text, Classification, Insert/edit |
| Jurassic-2 Mid | 8k | ai21.j2-mid-v1 | Jurassic-2 Mid is AI21’s mid-sized model, carefully designed to strike the right balance between exceptional quality and affordability. Jurassic-2 Mid can be applied to any language comprehension or generation task including question answering, summarization, long-form copy generation, advanced information extraction and many others. | Open book question answering, summarization, draft generation, information extraction, ideation | Text, Classification, Insert/edit, Math |
| Titan Text Express | 8k | amazon.titan-text-express-v1 | Amazon Titan Text Express has a context length of up to 8,000 tokens, making it well-suited for a wide range of advanced, general language tasks such as open-ended text generation and conversational chat, as well as support within Retrieval Augmented Generation (RAG). At launch, the model is optimized for English, with multilingual support for more than 100 additional languages available in preview. | Text generation, Code generation, Rich text formatting, Orchestration (Agents), Fine Tuning | Text generation, Code generation, Instruction following |
| Titan Text Lite | 4k | amazon.titan-text-lite-v1 | Amazon Titan Text Lite is a light weight efficient model ideal for fine-tuning for English-language tasks, including like summarization and copywriting, where customers want a smaller, more cost-effective model that is also highly customizable. | Text generation, Code generation, Rich text formatting, Orchestration (Agents), Fine Tuning | Fine Tuning, Text generation, Code generation, Rich text formatting |
As you can see, each model has different token size and they fit a different usecase; you may need to play with your hackathon product to find the right model for the job. When using the API, you will be able to change the FM that is in use by playing with the model_id.
NOTE: 1 token is equal to 3/4 of a word or 4 chars .
Why Bedrock and not chatGPT?
One of the biggest advantages of bedrock over other foundation models as chatGPT or Bard is that Bedrock allows us to access to multiple foundation models as noted above. Moreover, the most critical aspect that gave Bedrock winning edge is that it allows us to use our private data as context to the foundation model. More info on this can be found in the RAG part.