DaisyRose-27B: 1,200-task RL took Qwen3.8 from 5.6% to 12% on Terminal-Bench 4.1 RadixArk released DaisyRose-27B, its first model, built on Qwen3.8-27B and trained with reinforcement learning via the company's open-source Miles framework, lifting Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%. RadixArk trained DaisyRose-27B on 1,200 curated long-horizon agent tasks in 18 days on 144 H200 GPUs, with sessions up to 1M tokens, and released BF16, FP8 and NVFP4 checkpoints. RadixArk reported 37% model FLOPs utilization on its internal training backend versus 16% for the open-source Megatron-based trainer, cutting training cost 57% and reaching about 16,800 tokens per second end to end versus 8,300 for the open-source trainer. All posts https://www.radixark.com/blog DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency Today we are releasing DaisyRose-27B , RadixArk’s first model, built on Qwen3.8-27B and trained with reinforcement learning using Miles https://miles.radixark.com/ . We trained it on long-horizon agent tasks with sessions up to 1M tokens, with just 1,200 curated tasks in 18 days on 144 H200 GPUs. DaisyRose-27B improves Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%. Beyond Terminal-Bench 4.1, it also delivers gains on a variety of other reasoning, instruction following, and coding benchmarks. We are releasing BF16 , FP8 , and NVFP4 https://huggingface.co/RadixArk/DaisyRose-27B-NVFP4 checkpoints for everyone to try and build on. In this post, we explain how carefully designed tasks and efficient infrastructure made these gains possible on a limited compute budget. DaisyRose-27B scores higher than the original Qwen3.8-27B model on every benchmark above