Enterprise AI costs have become a boardroom obsession, and on Thursday, Writer responded with a new flagship model and a revamped harness designed to cut token spend dramatically. The company, which sells AI tools and agents for marketers, launched Palmyra X6, a post-training variation of Z.ai’s open-source GLM-5.2 model. Writer says the combination of the new model and upgrades to its agentic harness could reduce costs for customers by as much as 50% for basic tasks.
Both the model and the harness upgrades are available to Writer clients starting Thursday. The move comes as enterprises increasingly push back against the rising cost of AI deployments, with many CIOs questioning the value of proprietary models that demand premium per-token pricing.
Also read: OpenAI launches Ultrafast mode for GPT 5.6 Sol, claiming 14x speed boost
Why harness efficiency matters as much as model choice
Writer’s approach centers on the harness — the software layer that orchestrates how an AI model executes tasks, manages context, and calls tools. In a recent paper, Writer researchers tested small changes in harness efficiency across multiple models and found that harness tweaks were often a more reliable cost lever than switching models. Their testing showed cost reductions averaging 40%.
“The harness is the one component whose efficiency multiplies across every model an organization runs—present and future,” the researchers wrote. This insight is central to Writer’s pitch: rather than forcing customers to chase the latest benchmark-topping model, the company aims to optimize the infrastructure around the model.
Also read: IBM expands OpenAI partnership to push enterprise AI adoption
For complex, multi-step tasks — such as generating a marketing campaign or analyzing a sales pipeline — the harness can significantly reduce the number of tokens needed by avoiding redundant reasoning and streamlining tool calls. Writer’s upgrades focus on exactly these scenarios, making the system faster and cheaper for the workloads enterprises actually run.
Model-agnostic strategy and industry pushback
Writer’s platform remains model-agnostic. Palmyra X6 will sit alongside other Writer models or external models imported through Azure or Amazon Bedrock, giving clients flexibility. This approach reflects a broader industry trend toward hybrid deployments, where enterprises mix open-source and proprietary models based on cost and performance needs.
CEO May Habib framed the launch as a response to enterprise fatigue with the AI labs’ benchmark arms race. “I think the enterprise is absolutely sick of chasing the next benchmark,” she told TechCrunch. “They want flattening cost, and it seems like nobody can deliver that.”
Habib also suggested that rising costs are fueling distrust toward major AI labs, which have a financial incentive to drive up token usage. “The cost explosion here is just remarkable for customers, and so is the degree to which CIOs are giving up on the labs,” she said, adding that the labs “don’t deeply understand right how to help an enterprise get benefit from AI.”
This sentiment echoes a growing chorus of enterprise technology leaders who have publicly questioned the ROI of generative AI investments. A 2025 survey by Gartner found that nearly a third of enterprise AI projects were abandoned after proof of concept due to cost and complexity — a figure that has pushed vendors to emphasize efficiency over raw capability.
What this means for enterprise AI buyers
For enterprises, the launch signals a shift in how AI vendors compete. Instead of only touting benchmark scores, companies like Writer are now selling cost predictability and harness efficiency. That could pressure larger labs to offer more transparent pricing or optimization tools, especially as open-source models like GLM-5.2 continue to close the performance gap.
Writer’s bet is that enterprises will prioritize total cost of ownership over model supremacy. By offering a model-agnostic platform with a cost-optimizing harness, the company positions itself as a neutral intermediary — a stark contrast to labs that lock customers into proprietary ecosystems.
The next few months will show whether other vendors follow suit with similar cost-focused offerings. For now, Writer’s clients can start testing Palmyra X6 and the upgraded harness, with the promise of up to 50% savings on basic tasks.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. AI deployment costs and vendor claims can change rapidly; enterprises should conduct their own evaluation before making purchasing decisions.
Be the first to comment