Story

On air. The free wire runs 15 minutes behind live. See pricing

World, crypto, finance and tech news, scored for impact.

TechHacker NewsFiled 16 h ago

DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency

Open the sourceReported by Hacker News, radixark.com, 18 November 2026

Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.

RadixArk credits high-quality data, an asynchronous training setup (128 GPUs for inference, 16 for training) and robust sandboxes, claiming 57% lower training cost and 37% useful compute utilization. BF16, FP8 and NVFP4 checkpoints are being released for public use. The result matters as evidence that targeted post-training, not just raw compute, can move hard agentic benchmarks on a limited budget; watch whether the gains hold on independent evaluations and whether the checkpoint release spurs broader adoption of the Miles stack.

Muhabbit's AI wrote this summary in its own words from Hacker News's report. It can get details wrong, so check the source.

Named in this story

DaisyRose-27B, RadixArk, Qwen3.8-27B, Eigent AI, Miles