Writer launches Palmyra X6, a GLM-5.2-based model with harness upgrades to cut token costs
Writer introduced Palmyra X6, a post-training variant of Z.ai's open source GLM-5.2, alongside major upgrades to its agentic harness. The company estimates combined changes could cut customer costs by up to 50% for basic tasks, with harness efficiency alone averaging a 40% cost reduction across tests. The new model is deployment-ready at a lower price and available Thursday. CEO May Habib framed the move as addressing enterprises weary of chasing benchmarks and distrustful of major labs' incentives to drive token usage. This signals a shift toward infrastructure optimization as a key lever for cost control, rather than relying solely on model selection. A caveat: the 50% figure is an estimate and applies to basic tasks, while harness gains were measured across models in a recent Writer paper.
Writer launches Palmyra X6, a GLM-5.2-based model with harness upgrades to cut token costs
Writer launched Palmyra X6, a flagship model built on Z.ai's open source GLM-5.2, alongside upgrades to its agentic harness. The company estimates the combined changes will cut costs for customers by as much as 50% for basic tasks; both are available to Writer clients starting Thursday.
Key takeaway
Cost control is now an infrastructure problem, not just a model problem; harness efficiency multiplies across every model an organization runs.
What happened
Writer launched its flagship model Palmyra X6, built as a post-training variation on Z.ai's open source model GLM-5.2, alongside significant upgrades to its standard agentic harness. The company estimates the new system will cut costs for customers by as much as 50% for basic tasks, with both features available starting Thursday.
A recent paper from Writer researchers found that harness changes were a more reliable way to reduce costs than model choice, with costs falling an average of 40% across testing. CEO May Habib told TechCrunch that enterprises are 'absolutely sick of chasing the next benchmark' and that CIOs are giving up on the major AI labs.
Evidence
Writer's new Palmyra X6 is a post-training variation on Z.ai's open source model GLM-5.2.
TechCrunch AI · attributed
Built as a post-training variation on Z.ai's open source model GLM-5.2, Writer says the new system should provide deployment-ready capabilities at a much lower price.
The combined model and harness upgrades could cut costs by as much as 50% for basic tasks.
TechCrunch AI · attributed
The company estimates the new model, combined with changes to the companies harness infrastructure, will cut costs for its customers by as much as 50% for basic tasks.
Harness changes alone accounted for an average 40% drop in costs in Writer's research.
TechCrunch AI · attributed
The research found that, in many cases, changes in the harness were a more reliable way to reduce costs than model choice, with costs falling an average of 40% across their testing.
CEO May Habib says enterprises are 'absolutely sick of chasing the next benchmark' and wants flattening costs.
TechCrunch AI · attributed
I think the enterprise is absolutely sick of chasing the next benchmark. They want flattening cost, and it seems like nobody can deliver that.
Why it matters
For builders and operators, optimizing inference pipelines and harness layers is becoming a primary lever to control costs, as enterprises increasingly distrust major AI labs and demand flattening expenses.
Limits and uncertainties
The 50% cost reduction figure is an estimate and applies specifically to basic tasks, not all workloads.
The 40% average harness cost reduction comes from Writer's own paper, which may not generalize to all environments.
The article does not specify the exact token pricing or compare Palmyra X6 to GLM-5.2 directly on cost per token.
Practical implications
Enterprise builders should evaluate harness efficiency as a cost lever alongside model selection, not treat model choice as the only variable.
Organizations running multiple models may benefit from standardizing on a harness that multiplies efficiency across present and future models.
When considering Writer's offerings, customers should verify the cost savings on their specific task types, as the 50% claim is for basic tasks.
What to watch
Whether Writer's cost savings hold in real-world deployments and whether other vendors follow with similar harness-led cost reductions.
Adoption of Palmyra X6 by Writer clients and any performance benchmarks comparing it to base GLM-5.2.
Statements from major AI labs in response to growing enterprise distrust and cost pressure.
