Author Introduces Innovative AI Model and Advanced Harness for Managing Token Costs

As users in the AI industry become more mindful of the escalating costs tied to their deployments, there is a pressing need to curtail expenses. Although open-source models typically incur lower per-token charges, pinpointing the right model for particular tasks can prove difficult.

On Thursday, Writer, an AI tools and agents provider for marketers, launched its flagship model, Palmyra X6, aimed at alleviating this challenge for users. Created as a post-training variant of Z.ai’s open-source model GLM-5.2, Writer asserts that the new system will deliver deployment-ready capabilities at a notably reduced cost. The company anticipates that this model, along with infrastructure enhancements, could cut expenses for clients by as much as 50% for basic tasks.

Alongside the launch of the new model, the company revealed significant upgrades to its standard agentic harness. Both features will be accessible to Writer clients starting Thursday.

“I think enterprises are genuinely weary of chasing the next benchmark,” CEO May Habib shared with TechCrunch. “They seek predictable costs, and it appears that no one can provide that.”

This new approach places a strong emphasis on complex, multi-step tasks, executed more swiftly and with fewer tokens. Writer recognizes that optimizing the harness is crucial to realizing this goal.

A recent study by Writer’s researchers backs up this strategy, exploring minor tweaks in harness efficiency across diverse models. The results indicated that, in numerous cases, modifying the harness was a more reliable method of cost reduction than changing models, leading to an average cost decrease of 40% during their experiments.

“The harness is the singular component whose efficiency magnifies across every model an organization employs—both now and in the future,” the researchers noted.

For Writer’s clientele, the experience remains model-agnostic: Palmyra X6 will operate alongside other Writer models or third-party models integrated via Azure or Amazon Bedrock. However, Habib also observes that the drive for cost reduction is cultivating a rising skepticism toward large AI laboratories, which have a financial motive to boost token usage.

“The rapid rise in costs is unparalleled for clients, as is the level of distrust CIOs are developing toward the labs,” Habib stated to TechCrunch, adding that the AI labs “are currently not fully aware of how to enable enterprises to leverage AI effectively.”