Generative artificial intelligence startup Writer Inc. today released its newest state-of-the-art enterprise-focused large language model Palmyra X5, an adaptive reasoning model that features a 1 million-token context window.

The company also said that it teamed up with Amazon Web Services Inc. to announce the availability of the new model on Amazon Bedrock, AWS’s fully managed service for building AI applications by providing access to LLMs and machine learning models.

In an interview with SiliconANGLE, Matan-Paul Shetrit, director of product management at Writer, said Palmyra X5 was trained entirely using synthetic data and required just $1 million in graphics processing unit hours. “Writer has been a leader from very early on in the use of synthetic data, as well as techniques that allow us to cut down costs, like stopping training if we don’t see returns, which lets us push model pricing way down,” Shetrit said.

With its 1 million-token large-context window, the model can ingest and comprehend extensive codebases or large numbers of documents simultaneously. It also enables multistep reasoning across actions, making it ideal for powering agentic AI workflows.

“Palmyra X5 is the newest model that comes with a 1 million-token context window, which is one of the largest commercially available, and it’s amongst the fastest and cheapest large-context LLMs,” Shetrit said. “This is critical when it comes to building multi-agentic systems, where the responses need to be continuously fed back into the context window across multistep workflows.”

Shetrit said that going forward, Writer will incorporate the 1 million-token context window into all of Writer’s Palmyra family of models.

The model can read an entire 1 million-token prompt in about 22 seconds, and it can return individual function-calling turns in about 300 milliseconds. X5 performs with a pricing of 60 cents per 1 million input tokens and $6 per 1 million output tokens.

Between the low latency and low cost, it permits the model to provide high-speed performance across long contexts while allowing enterprise users to perform numerous actions without costing too much. This is essential, especially for agentic workflows, which can quickly become costly and slow with multistep and multiturn AI generations.

“If you’re thinking about enterprise use cases, they’re extremely complex and interacting with multiple third-party, external or internal systems, often in multistep flows,” Shetrit said. “Our focus isn’t on solving for one- or two-agent workflows. It’s on enabling enterprises to manage 10, 50, or even 100,000 agents all working alongside employees to drive real transformational change.”

The release of Palmyra X5 follows Writer’s launch of AI HQ, a new centralized hub for teams to easily build, deploy and orchestrate AI agents. Shetrit described Palmyra X5 and the company’s family of other Palmyra models as powerful “brains” for agents. He said X5 delivers on the company’s vision for helping companies adapt to the agentic AI trend.

“Palmyra represents a critical step forward for companies seeking to embed generative AI into their enterprise workflows,” said David Cushman, executive research leader at HFS Research Ltd. “Palmyra is purpose-built for agent development, API function calling, multimodal processing and custom enterprise AI applications. More importantly, it’s integrated into Writer’s end-to-end platform that balances scalability with transparency — a core demand from regulators and stakeholders alike.”

Hundreds of enterprise companies already use Writer’s Palmyra family of LLMs internally and externally to power their AI capabilities, including Accenture Plc, Marriott International Inc., Prudential Financial Inc., Qualcomm Inc., Uber Technologies Inc., Salesforce Inc. and The Vanguard Group Inc.

Image: Writer