ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)
RadixArk releases DaisyRose-27B, a 27-billion parameter model built on Qwen3.8-27B, improving Terminal-Bench 4.1 pass@1 scores from 5.6% to 12.0%. The team trained the model using 1,200 curated tasks over 18 days on 144 H200 GPUs, achieving 57% lower training costs and 2x faster end-to-end speeds compared to open-source trainers. The release includes BF16, FP8, and NVFP4 checkpoints available for download.
Tap to vote and see what everyone thinks.
Summary by ByteBrief