Writer has launched its flagship Palmyra X6 AI model alongside an upgraded, model-agnostic agent harness to address soaring enterprise token costs. Built on Z.ai's open-source GLM-5.2, Palmyra X6 combined with the new harness reduces basic task costs by up to 52% and cuts median execution times from 50 to 26 seconds. Writer CEO May Habib highlights a growing enterprise backlash against major AI labs, noting that businesses are tired of chasing benchmarks and want flattening, predictable costs.
Palmyra X6 model launch
- ▪The Palmyra X6 model is built as a post-training variation on Z.ai's open-source model GLM-5.2.
- ▪Palmyra X6 is priced at $2.00 per million input tokens and $8.00 per million output tokens, compared to Palmyra X5's pricing of $0.60 per million input tokens and $6.00 per million output tokens.
- ▪Writer launched its new flagship enterprise AI model, Palmyra X6, on August 13, 2026.
Enterprise AI cost reduction
- ▪Writer reported that its WRITER Agent product runs at a 52% lower cost per task when powered by Palmyra X6, reducing the price per task from $0.25 to $0.12.
- ▪Writer claims that combining Palmyra X6 with its upgraded agent harness infrastructure reduces the cost of completing basic enterprise tasks by up to 50%.
- ▪Under Palmyra X6, the median execution time fell from 50 seconds to 26 seconds per task, while average tokens used per task dropped by 38%.
Harness orchestration layer optimization
- ▪Writer released significant upgrades to its agent harness, which is the orchestration layer that manages requests, tool connections, and action sequences for AI agents.
- ▪A Writer research paper published on arXiv in July 2026 found that optimizing the harness layer alone reduced AI costs by an average of 40% across multiple models.
Model-agnostic efficiency gains
- ▪Writer's upgraded harness is model-agnostic, allowing it to work with Writer's proprietary models as well as third-party models connected through Microsoft Azure or Amazon Bedrock.
- ▪Writer's model-agnostic approach allows clients to scale orchestration efficiency gains across any model an organization runs, present and future.
Enterprise distrust of AI labs
- ▪Writer CEO May Habib stated that enterprise customers are tired of chasing model benchmarks and are instead demanding predictable, flattening costs.
- ▪Writer CEO May Habib asserted that enterprise buyers are growing distrustful of major AI labs because those labs have financial incentives to increase token consumption rather than help customers spend less.
Story comments
Loading comments…