Across the AI industry, users are becoming more aware of how much it costs to deploy, and there is a new sense of urgency to reduce costs. However, while open source models have a significantly lower cost per token, it can be difficult to find the right model for a specific job.
On Thursday, Writer, which provides AI tools and agents for marketers, launched a new flagship model called the Palmyra X6, aimed at solving user problems. Built as a post-training variation of Z.ai’s open source model GLM-5.2, the new system should offer deployment-ready functionality at a much lower price point, Writer said. The company estimates that the new model, combined with changes to the infrastructure used by enterprises, can reduce the cost of basic tasks for customers by up to 50%.
Along with the new model, the company has also released significant upgrades to its standard agent harness. Both features will be available in the Writer client starting Thursday.
“I think companies are completely tired of chasing the next benchmark,” CEO May Habib told TechCrunch. “They want cost levelization, but no one seems to be able to do that.”
The new approach focuses specifically on performing complex multi-step tasks faster and with fewer tokens. And Writer believes harness optimization is a key way to achieve that.
A recent paper by Writer researchers tests small changes in harness efficiency across several different models, lending credibility to this approach. The study found that harness changes were often a more reliable way to reduce costs than model selection, with an average cost reduction of 40% across tests.
“Harnesses are one component that multiplies efficiency across all current and future models that organizations run,” the researchers wrote.
For Writer’s clients, the experience remains model agnostic. Palmyra X6 works alongside other Writer models and external models imported via Azure or Amazon Bedrock. However, Habib also sees the drive to cut costs as leading to a growing distrust of large AI labs, which have economic incentives to promote the use of tokens.
“The cost explosion here is truly unprecedented for customers, and so is the degree to which CIOs are abandoning labs,” Habib told TechCrunch, adding that AI labs “don’t really understand how to help companies reap the benefits of AI.”
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
