Writer introduces new AI model and upgraded harness to contain token costs
Writer says Palmyra X6 and harness upgrades can cut basic-task costs for customers by up to 50%.
The new model is a post-training variation on Z.ai’s open source GLM-5.2, built to sit alongside Writer’s own models and outside models brought through Azure or Amazon Bedrock. Writer is also upgrading its agentic harness, arguing that harness efficiency can reduce token use across complex, multi-step tasks. The company points to its own research finding harness changes cut costs by an average of 40% in testing. CEO May Habib framed the launch as a response to enterprises tired of benchmark-chasing and rising deployment costs. TechCrunch AI's note
The new model is a post-training variation on Z.ai’s open source GLM-5.2, built to sit alongside Writer’s own models and outside models brought through Azure or Amazon Bedrock. Writer is also upgrading its agentic harness, arguing that harness efficiency can reduce token use across complex, multi-step tasks. The company points to its own research finding harness changes cut costs by an average of 40% in testing. CEO May Habib framed the launch as a response to enterprises tired of benchmark-chasing and rising deployment costs. TechCrunch AI's note
score 5