RELEASE · MODELS · #3867
RadixArk releases DaisyRose-27B, a Qwen3.8-27B-based RL-post-trained 27B model
RadixArk released DaisyRose-27B, a 27B-parameter model built on Qwen3.8-27B and post-trained with reinforcement learning using their Miles framework on long-horizon agent tasks (sessions up to 1M tokens). Trained over 18 days on 144 H200 GPUs using ~1,200 curated tasks, the model raises Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%; checkpoints are available under Apache 2.0 and the team recommends serving with SGLang and DSpark speculative decoding.
KEY POINTS
- RadixArk released DaisyRose-27B, a 27B-parameter model built on Qwen3.8-27B and post-trained with reinforcement learning using their Miles framework on long-horizon agent tasks (sessions up to 1M tokens).
- Trained over 18 days on 144 H200 GPUs using ~1,200 curated tasks, the model raises Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%; checkpoints are available under Apache 2.0 and the team recommends serving with SGLang and DSpark speculative decoding.
- This release demonstrates that targeted post-training with long-horizon agent tasks and an asynchronous RL pipeline can substantially improve a base model's benchmark performance and cut GPU training time (reported 57% lower cost), while publishing permissively licensed checkpoints.
WHY IT MATTERS
This release demonstrates that targeted post-training with long-horizon agent tasks and an asynchronous RL pipeline can substantially improve a base model's benchmark performance and cut GPU training time (reported 57% lower cost), while publishing permissively licensed checkpoints.