Enterprise AI costs are climbing fast, and users are pushing back. On Thursday, Writer launched a new flagship model called Palmyra X6, built as a post-training variation on Z.ai's open source model GLM-5.2. The company says the model, combined with upgrades to its agentic harness infrastructure, could cut customer costs by as much as 50% for basic tasks.
Writer sells AI tools and agents for marketing teams. The new model and harness upgrades are available to clients starting Thursday.
A shift away from benchmark chasing
CEO May Habib said enterprises are tired of paying a premium for incremental model improvements. The new approach focuses on complex, multi-step tasks executed faster with fewer tokens, with harness optimization as the primary lever.
"I think the enterprise is absolutely sick of chasing the next benchmark," Habib told TechCrunch. "They want flattening cost, and it seems like nobody can deliver that."
A research paper from Writer engineers tested small efficiency changes in harnesses across several models. They found harness tweaks reduced costs by an average of 40%, often more reliably than switching models.
"The harness is the one component whose efficiency multiplies across every model an organization runs-present and future," the researchers wrote.
A maturing market for AI agents
Writer's clients can still choose between Palmy X6, other Writer models, or outside models imported through Azure or Amazon Bedrock. The company's core pitch: lower cost driven by better infrastructure, not just a faster model.
Habib also sees cost pressure pushing enterprises away from major AI labs that profit from heavier token usage. "The cost explosion here is just to customers, and so is the degree to which CIOs are giving up on the labs," Habib said. "They don't deeply understand right now how to help an enterprise get beenefit from AI."
As teams weigh options between open source per-token pricing and proprietary convenience, the harness approach offers a practical middle ground. AI agents and automation training can help teams get familiar with how these systems work before they buy in.
Why this matters for writers
Writers who rely on AI tools are down - a 50% cut in token costs changes the math on how much output you can afford per assignment. If you test models for drafting, revision, or research workflows, the cost saving matters. For those who use writer-specific platforms or invest in general tools, keeping an eye on per-token pricing trends in your existing stack helps whether you stay or switch.
Your membership also unlocks:
