A Guth Labs publication ↗
Back to News

OpenRouter adds RadixArk’s DaisyRose 27B model

A faceted cream ceramic flower with a gold center and sage paths sits on a glass base.
AI-generated editorial illustration

OpenRouter added RadixArk’s DaisyRose 27B to its catalog on November 21, giving developers a hosted API for an open-weight model that was already available to download. OpenRouter lists DeepInfra as the serving provider, with listed prices of $0.15 per million input tokens and $1.875 per million output tokens. OpenRouter marks those rates as 25% off.

DaisyRose accepts text and images and returns text, with a 262,144-token context window on OpenRouter. Reasoning is enabled by default and can be turned off. OpenRouter says the model is not currently available through its EU in-region routing option.

RadixArk built DaisyRose from Qwen3.8 27B and describes the training run as reinforcement learning on long-running agent tasks, without supervised fine-tuning or distillation. The downloadable weights remain available under Apache 2.0. The new listing gives teams a way to test the same model without running it themselves. In return, teams pay for the tokens their applications use and depend on OpenRouter’s currently listed provider.

Sources and citations

Revision history

Published .