Writer, a company that provides AI tools and agents for marketers, launched a new flagship model called Palmyra X6 on Thursday, promising to slash costs for its customers by as much as 50% for basic tasks. The new system, built as a post-training variation on Z.ai's open source model GLM-5.2, aims to deliver deployment-ready capabilities at a significantly reduced price point. Combined with substantial upgrades to Writer's standard agentic harness, the release addresses mounting pressure from enterprise clients to flatten AI expenses.
The cost reduction comes from two sources working in tandem. Writer estimates that Palmyra X6, together with changes to the company's harness infrastructure, will cut expenses by up to 50% for basic tasks. A recent paper from Writer researchers tested minor adjustments in harness efficiency across several different models and discovered that harness modifications were often a more dependable method to lower costs than switching models, with expenses dropping an average of 40% across their testing. The new approach emphasizes complex, multi-step tasks that can be executed more quickly and with fewer tokens.
"I think the enterprise is absolutely sick of chasing the next benchmark," CEO May Habib told TechCrunch. "They want flattening cost, and it seems like nobody can deliver that." The researchers wrote that "the harness is the one component whose efficiency multiplies across every model an organization runs—present and future," underscoring why Writer sees harness optimization as essential. For Writer's clients, the experience remains model-agnostic: Palmyra X6 will sit alongside other Writer models or outside models imported through Azure or Amazon Bedrock. Both features became available to Writer clients starting Thursday.
Habib also sees the push to cut costs as driving a broader distrust toward major AI labs, which have a financial incentive to drive up token use. The CEO told TechCrunch that "the cost explosion here is just unprecedented for customers, and so is the degree to which CIOs are giving up on the labs," adding that the AI labs "don't deeply understand right how to help an enterprise get benefit from AI." While open source models offer substantially lower per-token costs, finding the right model for a specific job can be challenging. Writer's strategy addresses this by optimizing the infrastructure layer that sits between enterprises and whatever models they choose to run, making efficiency gains portable across present and future deployments.
The release signals Writer's bet that enterprise customers care more about controlling expenses than accessing the highest-performing models available. By focusing on harness efficiency rather than benchmark performance, the company positions itself as a cost-management layer for organizations running AI at scale. The approach could reshape how enterprises evaluate AI vendors, shifting the conversation from capability comparisons to total cost of ownership. If Writer's research holds across broader deployments, infrastructure optimization may become the primary battleground for AI platform providers, rather than the model performance race that has dominated the industry's attention. For CIOs managing ballooning AI budgets, the question may no longer be which model performs best, but which vendor can make existing models run leaner.