DaisyRose-27B was developed through an 18-day reinforcement learning run on 144 H200 GPUs using Qwen3.8-27B as its foundation. The training leveraged 1,200 curated tasks created with Eigent AI across science, machine learning, and software engineering, requiring models to navigate chained dependencies and up to 1.5 hours of sandbox runtime. RadixArk supported the training effort using an internal sandbox system built on AgentENV that executed roughly 944,000 sandboxes between mid-September and mid-November.
Key points
The internal training backend reached 37% model FLOPs utilization, cutting the necessary…: The internal training backend reached 37% model FLOPs utilization, cutting the necessary GPU time by 57% compared to Megatron-based alternatives.
The 63,000 GPU-hour run processed roughly 16,800 tokens per second, delivering an end-to-…: The 63,000 GPU-hour run processed roughly 16,800 tokens per second, delivering an end-to-end speed twice as fast as open-source trainers.
Public checkpoint releases in BF16, FP8, and NVFP4 formats expand access to models tuned…: Public checkpoint releases in BF16, FP8, and NVFP4 formats expand access to models tuned for long-horizon agent execution.
Sources checked by our analysis pipeline: 1
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. 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 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 above1. The largest gain is on Terminal Bench 4.1, a notoriously hard long-horizon benchmark. Our magic lies in the following three key ingredients: high-quality training data, efficient training backend, and robust sandboxes.
High-quality Training Data
We built the tasks in partnership with Eigent AI, starting from Eigent AI’s large pool of tasks in machine learning, software…
Addressing widespread workforce anxiety surrounding automated systems, leadership offered voluntary buyouts to tens of thousands of staff to facilitate an internal reorganization around artificial intelligence. The strategy grants employees personal autonomy during sweeping workflow changes rather than imposing forced terminations. Ultimately, management expects all internal roles to transition from manual execution to directing and overseeing autonomous digital tools.
Capital secured from backers in Japan marks a major milestone for the artificial intelligence company as it completes its public listing. However, the final proceeds settled at the bottom of the targeted expectations outlined in earlier projections. The resulting total reflects the minimum threshold management had sought to achieve during this institutional offering phase.
The aggressive buildout of artificial intelligence facilities is placing unprecedented strain on natural resources and local municipal budgets across the country. Power providers and military developers are facing intense pushback over fossil fuel expansion and resource consumption. At the same time, communities and educational institutions are grappling with the societal friction caused by imperfect automated tools.