ZeroDailyGet it on Google Play

DaisyRose-27B: A Leap in Post-Training Efficiency

Original headline: DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency — huggingface.co

More in AI and machine learning news

RadixArk has released DaisyRose-27B, its first model, built on Qwen3.8-27B and fine-tuned using reinforcement learning. Trained on just 1,200 curated tasks over 18 days on 144 H200 GPUs, it handles long-horizon agent tasks with sessions up to 1 million tokens. The model retains Qwen3.8-27B's core architecture, including its 256K-token context window and image understanding capabilities.

For developers, this means a significant efficiency boost in post-training—doubling performance on Terminal-Bench 4.1 from 5.6% to 12.0%—while maintaining the original model's versatility. Available under Apache 2.0, DaisyRose-27B demonstrates how focused reinforcement learning can extract more capability from existing architectures without starting from scratch.

Read the original at huggingface.co

Satirical commentary on real tech news, summarised with AI. Read the original source for the facts. How ZeroDaily works

Open in the appGet it on Google Play

More stories

Add ZeroDaily on Google