344
Productivity & Workflow355
Automation & Workflow224
Software Development251
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics142
Photography & Imaging156
Design & Creative170
Customer Support132
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
Writer introduced its Palmyra X6 AI model and upgraded its agent harness, saying the changes could cut some customer AI costs by up to 50%.
In short: Writer released a new AI model called Palmyra X6 and upgraded its agent harness, aiming to lower AI running costs for business customers.
Writer, a company that sells AI tools for marketing and other business work, announced a new flagship AI model named Palmyra X6. The company says the model is designed to be cheaper to use in real products and workflows.
Palmyra X6 is based on an open source model from Z.ai called GLM-5.2. Open source here means the basic model design is shared publicly, so other companies can build on top of it.
At the same time, Writer upgraded its “agentic harness,” which is the software layer that helps an AI system carry out tasks step by step (like a set of instructions and tools around the model). Writer estimates that Palmyra X6 plus these harness changes can cut customer costs by as much as 50% for basic tasks.
Writer said both the new model and the harness upgrades were available to its clients starting Thursday. The company also said customers can still use other Writer models, or bring in outside models through services like Microsoft Azure or Amazon Bedrock.
Many companies pay for AI based on “tokens,” which are small chunks of text, similar to paying per word. If an AI takes more steps and produces more text than needed, the bill can climb quickly.
Writer is betting that improving the harness can reduce wasted steps and text, especially on multi-step tasks. In a recent research paper, Writer researchers reported that small harness efficiency changes lowered costs by an average of 40% across their tests, sometimes more reliably than switching to a different model.
Source: TechCrunch AI