Enterprise AI platform WRITER has launched Palmyra X6, a new flagship model designed to improve the economics, speed, and reliability of agentic AI workflows used by marketing and revenue teams. The company also introduced major upgrades to its WRITER Agent orchestration system alongside expanded governance tools for monitoring adoption, performance, and token spending. WRITER said the combined releases are intended to help enterprises deploy more complex AI agents in production without rapidly rising usage costs limiting adoption.
Reducing the Cost of Enterprise AI Agents
The launch addresses a growing challenge for businesses moving from conversational AI tools toward autonomous agents capable of executing multi-step tasks. Such systems may reason through lengthy workflows, interact with enterprise tools, delegate work, and continuously process context, resulting in substantially higher token consumption than conventional chatbot interactions. According to WRITER's internal testing, WRITER Agent paired with Palmyra X6 reduced average costs by 52%, increased speed by 48%, and improved quality by 10%.
WRITER Chief Technology Officer and co-founder Waseem AlShikh said enterprises increasingly want employees to use AI extensively but need the underlying economics to remain manageable as adoption expands. He said the company has spent several years focusing on efficiency, reliability, and governance alongside model performance, particularly for organizations deploying AI in regulated environments. The latest releases are designed to make larger-scale agent deployments financially practical while preserving the control businesses require.
Palmyra X6 Targets Enterprise Workflows
Palmyra X6 was developed specifically for use within the WRITER Agent environment, with an emphasis on marketing and revenue-related work. Rather than relying exclusively on common public benchmarks, WRITER created internal evaluations based on production workflows used by its customers, assessing areas such as grounding, retrieval, tool use, content generation, sub-agent delegation, and brand voice. The model was post-trained on GLM 5.2 and evaluated across nine enterprise-oriented capabilities.
WRITER reported that X6 achieved an average score of 0.87 out of 1.00 across those evaluations, with pricing of $2 per million input tokens and $8 per million output tokens. The company said the result represented the strongest combination of performance and cost among the models it tested, which also included systems from Anthropic, OpenAI, and Google. WRITER further reported that X6 completes tasks in an average of 26 seconds, generates 82 tokens per second, and can work toward a single objective for as long as eight hours.
WRITER Agent Gains Faster Orchestration
Alongside the model release, WRITER rebuilt significant parts of its Agent harness to reduce unnecessary reasoning and token usage during complex workflows. The system can now adjust its reasoning strategy according to task complexity, directly answering simpler requests while generating structured execution plans for more demanding jobs. It can also process work in batches or delegate tasks to sub-agents, reducing repeated context and helping larger workflows run more efficiently.
Research published by WRITER found that the updated harness improved results even when third-party models were used instead of the company's own technology. Across the models tested, WRITER said tasks were completed 44% faster while average cost per task fell 41%, without a decline in overall quality. Multi-model support has also been extended to WRITER Agent, allowing administrators to enable models from providers including Anthropic and OpenAI or connect models hosted through Microsoft Azure, AWS Bedrock, and NVIDIA NIM.
Governance Tools Expand Cost Controls
WRITER also introduced additional reporting and governance capabilities aimed at giving enterprise leaders greater visibility into how agentic AI is being used across their organizations. Administrators can monitor adoption, performance, and spending while analyzing individual Playbooks and Skills, which are reusable workflows and instructions created within the platform. New consumption controls additionally allow companies to establish alerts and limits intended to prevent unexpected AI spending as usage increases.
The governance features are designed to help organizations identify frequently used workflows and determine which applications should be expanded across additional teams. H&R Block marketing operations executive Robyn Murphy said stronger governance had helped the company move AI beyond limited pilot programs, while lower task costs could make previously uneconomical high-volume workloads practical to automate. WRITER said all of the newly announced model, agent, and governance capabilities are available to customers immediately.
WRITER's latest release reflects a broader shift in enterprise AI from experimentation toward the economics of running autonomous systems at scale. By combining Palmyra X6 with a more efficient orchestration layer and stronger spending controls, the company is positioning cost efficiency and governance alongside raw model capability as critical factors in enterprise adoption. As businesses expand the use of agents across production workflows, controlling token consumption while maintaining reliability could become an increasingly important competitive differentiator for enterprise AI platforms.