Writer Palmyra models
The writer Palmyra X5 and X4 foundation models are now accessible on Amazon Bedrock. The growth of context windows in foundation models (FMs) is one development that has taken place in recent months; several of them can now handle sequence lengths that were unthinkable only a year ago. It is still difficult to create AI-powered apps that can handle enormous volumes of data while upholding the security and dependability requirements needed for business use.
Because of these factors, one can are pleased to inform that the fully managed, serverless Writer Palmyra X5 and X4 variants are now available on Amazon Bedrock. The first significant cloud provider to offer fully managed Writer models is AWS. Today, Writer unveiled a new model called the Palmyra X5. Previously, the Palmyra X4 was offered on Amazon Bedrock Marketplace.
Strong reasoning capabilities provided by Writer Palmyra models enable intricate agent-based workflows while upholding organizational security norms and dependability. Palmyra X4 offers a 128K token context window, while Palmyra X5 has a one million token context window. Deeper analysis and more thorough task completion are made possible by these models’ removal of some of the conventional limitations on app and agent development due to their large context windows.
The most cutting-edge models and the resources required to create generative AI applications with security, privacy, and ethical AI are now available with Amazon Bedrock.
Writer, a leader in FM creation, uses Amazon SageMaker HyperPod to train and optimize its market-leading models. Writer speeds up the release of its models by cutting down on training time with to its well-designed distributed training environment.
A wide range of enterprise-grade features are supported by Palmyra models, including:
- Adaptive thinking is a type of hybrid model that excels at addressing complicated problems and making complex decisions by fusing enterprise-grade dependability with sophisticated reasoning.
- Support for sophisticated tool-calling features that can be applied to intricate multistep workflows and agentic actions, such as interacting with corporate systems to carry out operations including sending emails, updating systems, conducting transactions, and initiating workflows.
- corporate-grade reliability: Models are explicitly trained on business material to align outputs with professional standards, ensuring consistent, accurate results while upholding the stringent quality standards necessary for corporate use.
Writer’s Palmyra in Amazon Bedrock
Use Palmyra X hybrid models to strike a balance between enterprise-grade dependability and sophisticated reasoning capabilities. Perform exceptionally well on intricate agentic workflows.
The Palmyra X5
Writer’s most sophisticated model, Palmyra X5, was created specifically for creating and expanding AI agents throughout the company. With the help of a unique transformer design and hybrid attention techniques, it provides industry-leading speed and efficiency on context windows up to one million tokens. Scaling AI agents requires faster inference and more memory to handle massive amounts of enterprise data and tool calls. In addition to supporting code generation, structured outputs, and more than 30 languages, Palmyra X5 has adaptive reasoning that allows it to dynamically modify approach based on circumstances.
The Palmyra X4
Palmyra X4, which ranks first on Stanford HELM, performs exceptionally well on challenging jobs and agentic workflows. It combines a number of features, such as multi-modality, multilingual support, built-in RAG, sophisticated reasoning, tool calling, large language models (LLMs) delegation, code generation, and structured outputs, with a 128K token context window. Developers may create apps and agents that update systems, complete transactions, send emails, initiate workflows, and more with Palmyra X4’s enterprise-specific tools, which expand the model’s actionable capabilities.
Advantages
Token context windows of 1M and 128K
1M and 128K tokens are processed by Palmyra X5 and X4, respectively. Processing large documents, intricate multi-step reasoning, and advanced agent-based workflows are all made possible by this enlarged window. Deeper analysis and more thorough job completion are made possible by X5’s 1M token context window, which successfully eliminates conventional context limits.
Adaptive reasoning
The hybrid Palmyra X5 and X4 models blend dependability with sophisticated reasoning skills. These models do exceptionally well in multi-step logical reasoning, complicated problem solving, and complex decision-making. The models can maintain consistent, accurate outputs across a range of use cases, comprehend complex context, and adjust to various business conditions.
Assist with agentic processes
Complex multi-step processes and agentic behaviors are made possible by these models’ support for sophisticated tool-calling capabilities. To carry out operations like system updates, transaction execution, email sending, and process initiation, the models can communicate with corporate systems. This feature makes it possible to develop intelligent agents and complex automated processes that can manage intricate commercial operations.
Enterprise-level dependability
Palmyra models perform well in practical commercial applications and are highly scored on industry-leading benchmarks such as Stanford’s HELM. While upholding the exacting quality requirements necessary for commercial use, they consistently produce accurate results. Because the models are trained on business material, the results are guaranteed to meet both business needs and professional standards.
Transparency and security
With its SOC 2 Type II, PCI, and HIPAA certifications, Writer offers superior security. Palmyra models guarantee total data privacy and sovereignty by not storing or using client data for model training. Businesses can use cutting-edge Artificial Intelligence capabilities while maintaining complete control over their data and putting their own security measures in place.
Palmyra X5 and X4 use cases
Services related to finance
Investment banking and asset and wealth management tools, such as deal transaction support, 10-Q/10-A and earnings transcript highlights, fund and market research, and mass customized customer outreach, are powered by Palmyra models.
Medical care and biological sciences
In order to develop solutions for member acquisition and onboarding, appeals and grievances, case and utilization management, employer RFP response, and other areas, payors and providers utilize Palmyra models. Pharmaceutical companies provide solutions for clinical trials, R&D, commercial medical affairs, and other areas using Palmyra models.
Consumer goods and retail
Palmyra models are used by retailers for a variety of purposes, including RFP research and response, brand and compliance assessments, performance analysis, SEO upgrades, and web page design and variation.
Technology
Personalized and account-based marketing, content production, campaign workflow automation, account preparation and research, knowledge support, job briefs and candidate reports, RFP responses, and more are all supported by Palmyra models, which also support AI solutions across go-to-market strategies, customer support, and operational functions.