Writer, a San Francisco-based AI startup, launched a new large language model on Wednesday designed to compete with offerings from tech giants like OpenAI and Anthropic. The model, called Palmyra X 004, cost approximately $700,000 to train, significantly less than the estimated $4.6 million for a similarly sized model from OpenAI. Writer is currently seeking up to $200 million from investors, which would bring its valuation to $1.9 billion, nearly four times its valuation from September 2023 when it raised $100 million at a valuation exceeding $500 million.
The company achieves lower training costs by using synthetic data – AI-generated data that mimics real-world information. This approach preserves privacy and is becoming increasingly popular for model training. “We don’t train our models on fake or hallucination data, and we don’t use a model to generate random data,” clarified Writer’s co-founder and CTO, Waseem Alshikh.
“We take real, factual data and convert it to synthetic data that is specifically structured in a clearer and cleaner way for model training.”
Writer’s generative AI enables corporate clients to generate human-like text for various applications, including social media posts, job descriptions, and mission statements. It can also analyze data, summarize text, and build custom AI applications for tasks like market analysis. The company has over 250 enterprise customers across sectors such as support, IT, operations, sales, and marketing.
Writer’s competitive AI model launch
Palmyra X 004 particularly excels in function calling and workflow execution, key capabilities for building practical AI agents and assistants. It achieved a score of 78.76% on a benchmark evaluating its ability to select appropriate tools, determine which APIs to call, and successfully execute tasks based on natural language inputs.
This surpassed offerings from major tech companies by nearly 20%. The model supports multilingual capabilities across 30+ languages and can handle multimodal inputs including text, images, and audio. Writer offers multiple deployment options, allowing companies to access the model through cloud providers or host it on-premises within their own infrastructure.
“We’re seeing a transition from using AI for simple tasks like summarizing emails to building complex, multi-step workflows,” noted Alshikh. “Our enterprise customers are looking to create AI agents that can interact with multiple internal systems, access varied data sources, and execute sophisticated business logic.”
As AI becomes more integrated into business processes, issues of reliability, explainability, and governance are paramount. Palmyra X 004 includes built-in features like automatic data integration and AI guardrails to allow enterprises to set content policies and control the model’s outputs.
Looking ahead, Writer is exploring ways to build even deeper transformer models with potentially 500-2000 layers, which they believe could lead to significant improvements in reasoning capabilities. The generative AI market is expected to see substantial growth in the coming years, with investors funneling billions into AI companies in 2023 and 2024.
