Writer, a provider of AI tools and agents for marketers, announced on Thursday a new flagship model named Palmyra X6 and significant upgrades to its standard agentic harness. The company says the combined changes are intended to reduce token costs for its customers.
Palmyra X6 is a post-training variant built on Z.ai’s open-source model GLM-5.2; Writer describes the approach as delivering deployment-ready capabilities at a lower price point. The company estimates that the new model together with changes to its harness infrastructure could cut customer costs by as much as 50% for basic tasks.
The update places special emphasis on complex, multi-step tasks, aiming to complete them faster and with fewer tokens. Writer considers harness optimization a crucial lever to achieve those efficiency gains, arguing that harness improvements multiply across every model an organization uses, both present and future.
A recent paper by Writer researchers supports this emphasis: small changes in harness efficiency tested across multiple models led to cost reductions averaging 40% in their experiments. The researchers concluded that in many cases harness modifications were a more reliable route to cost savings than switching models.
For Writer’s clients, Palmyra X6 will be available in a model-agnostic environment: it will sit alongside other Writer models and can be used together with external models imported via Azure or Amazon Bedrock. At the same time, CEO May Habib said the drive to cut costs is feeding broader distrust of large AI labs, which she argued have financial incentives to increase token use.
"The cost explosion here is just unprecedented for customers, and so is the degree to which CIOs are giving up on the labs," Habib told TechCrunch, adding that the AI labs "don’t deeply understand right how to help an enterprise get benefit from AI."
Writer stated that both Palmyra X6 and the harness upgrades became available to clients starting Thursday. The announcement highlights that reducing the cost of AI deployments may rely as much on optimizing the surrounding infrastructure as on selecting different models.
Key facts
- Product: Palmyra X6 (post-training variant of Z.ai’s GLM-5.2) and harness upgrades.
- Timing: announced and made available to customers on Thursday.
- Estimated impact: up to 50% cost reduction for basic tasks according to Writer; internal tests showed harness changes reduced costs by an average of 40%.
The release underlines a growing focus in the industry on practical cost savings and suggests harness-level improvements can be an effective complement to open-source model adoption.
