Writer launched flagship enterprise AI model Palmyra X6 on August 6, 2026, pitching lower deployment costs for companies using AI agents. Writer said benchmarked WRITER Agent task cost fell 52% to $0.12 from $0.25, median execution time dropped to 26 seconds from 50 seconds, output quality improved 10%, and average tokens per task fell 38%; in broader customer framing, it said the model and harness changes can cut basic-task costs by as much as 50%. Palmyra X6 is a post-training variation of Z.ai's open-source GLM-5.2, and Writer released upgraded agent harness software alongside it, with the model available next to other Writer models and outside models imported through Azure or Amazon Bedrock. CEO May Habib said enterprises are "absolutely sick of chasing the next benchmark" and want flatter AI costs, while a July 2026 arXiv paper from Writer researchers said harness-level optimization reduced costs by an average of 40% across tests and was often a more reliable lever than model choice alone. Writer also introduced governance controls to cap token use; although Palmyra X6 carries higher list prices than Palmyra X5 at $2.00 per million input tokens and $8.00 per million output tokens versus $0.60 and $6.00, the company says lower token consumption can still reduce effective output costs.