Writer launches Palmyra X6 to cut enterprise AI costs

Writer has launched Palmyra X6, its flagship artificial intelligence model, alongside upgrades to its agent platform and new governance tools aimed at lowering the cost of running AI agents in enterprise settings.

The San Francisco-based group designed the new model for marketing and revenue workflows, while the updated Writer Agent system is intended to make longer, multi-step tasks faster, cheaper and more reliable. It also introduced reporting and spending controls for administrators overseeing AI use across teams.

The move reflects a broader shift in the enterprise AI market, where attention has moved from experimentation to the economics of deployment. As businesses put AI agents into day-to-day operations, token consumption and inference costs have become a bigger concern than in earlier chatbot-style applications.

Writer said its agent system now runs at an average 52% lower cost, with a 48% improvement in speed and a 10% improvement in quality when paired with Palmyra X6. According to the company, the model can work on a single goal for up to eight hours while maintaining coherent reasoning across extended tasks.

Writer said Palmyra X6 was post-trained on GLM 5.2 and tested against several competing models using internal measures tied to production workflows. Those assessments covered grounding and retrieval, tool use, content generation, sub-agent delegation and brand voice.

On those tests, X6 achieved an average score of 0.87 out of 1.00 across nine evaluations, ahead of Claude Opus 4.8 at 0.86, Claude Sonnet 4.6 at 0.85, GPT-5.5 at 0.80 and Gemini 3.1 at 0.77. Writer priced the model at USD $2 per million input tokens and USD $8 per million output tokens, compared with USD $15 and USD $75 for Claude Opus 4.8, USD $3 and USD $15 for Claude Sonnet 4.6, USD $5 and USD $15 for GPT-5.5, and USD $2.50 and USD $10 for Gemini 3.1.

Writer said the model completes tasks in 26 seconds on average and generates 82 tokens per second. It also said X6 was the least politically biased model in its testing, which it described as important for teams producing customer-facing content under a corporate brand.

Agent changes

Alongside the model release, Writer rebuilt the backend of its agent harness to adjust reasoning based on the task. The system can answer simpler questions directly while creating more structured plans for complex work.

It now also supports batch execution and the use of sub-agents for higher-volume tasks. Writer said the changes are intended to reduce unnecessary steps and repeated context, which can increase both cost and the risk of workflows breaking down.

Citing its own research on orchestration design and token economics, Writer said the upgraded harness improved performance across both its own and third-party models. On average, the system completed tasks 44% faster and at a 41% lower cost per task while maintaining quality.

As part of the update, Writer Agent now includes extended multi-model support. Administrators can enable models from providers including Anthropic and OpenAI, while customers can also bring in their own models from Microsoft Azure, AWS Bedrock and NVIDIA NIM.

Controls and reporting

Writer also introduced new tools for monitoring adoption, performance and spending across an organisation. The additions give administrators a central view of how Writer Agent is being used, with filters to compare usage over time.

The reporting layer also adds analytics and governance for Playbooks and Skills, Writer's terms for repeatable agent workflows and instructions. Administrators can track usage, manage ownership, and review performance and costs at the level of individual workflows.

Consumption controls, including alerts and limits, have also been added to help manage spending across teams. Writer said the aim is to help leaders identify the most widely used workflows and expand those that show the strongest results.

Writer, founded in 2020, sells AI tools to large companies, particularly in marketing, sales and other business functions. It said all of the new features are now available to customers.

"The enterprise wants token consumption to explode - it means adoption is happening - but they need costs to flatten," said Waseem AlShikh, Chief Technology Officer and Co-Founder at Writer.

"For the past five years, Writer has focused on delivering not just frontier-level model performance, but the efficiency, reliability and governance that highly regulated enterprises need to confidently put agents into production. With Palmyra X6 and major harness improvements to Writer Agent, we're delivering a step change in both agent capabilities and cost that fundamentally changes how broadly and boldly our customers can deploy agentic AI," AlShikh said.