RadixArk releases DaisyRose-27B post-training model
11/19/2026 — 11/19, 04:21·1 sources (all)·1 reports (all)
Key takeaways
- RadixArk released its first model, DaisyRose-27B, built on Qwen3.8-27B and post-trained with reinforcement learning via Miles. Training used…
Background & analysis
On November 19, 2026, the Hugging Face Blog announced that RadixArk has released DaisyRose-27B, its first model, positioned around post-training efficiency. The model is built on Qwen3.8-27B and is trained with reinforcement learning using Miles, with a focus on long-horizon Agent tasks that support sessions of up to 1M tokens. Training relied on just 1,200 curated tasks and ran for 18 days on 144 H200 GPUs.
According to the official release, DaisyRose-27B lifts Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%, and also reports gains across a range of other reasoning, instruction following, and coding benchmarks. The model keeps the architecture, tokenizer, chat template, image understanding, and 256K token context window of Qwen3.8-27B. For longer sessions, harness summarization is used to connect multiple trajectories. The checkpoints are offered under the Apache 2.0 license, and the model card lists model properties and inference settings.
So far, public coverage of this release consists of the launch announcement alone, with no follow-up developments reported beyond it.
AI-generated from 1 reports · updated 3 days ago
Related AI tools
- Heat index
- 25
- All sources
- 1
- All reports
- 1
- First seen
- 3 days ago
Reports on this story headlines open the original
RadixArk released DaisyRose-27B, its first model, built on Qwen3.8-27B and trained with reinforcement learning via Miles on long-horizon agent tasks, using 1,200 curated tasks over 18 days on 144 H200 GPUs with sessions up to 1M tokens. The company reports Terminal-Bench 4.1 pass@1 rising from 5.6% to 12.0%, plus gains on other reasoning, instruction-following and coding benchmarks. Checkpoints are Apache 2.0, keeping the 256K context and image understanding.
Hugging Face BlogFirst-partyAI score 47
Other stories people are talking about
- Heat index 85Together AI offers Qwen4 Max at 30% off, fixing a Tornado bug in one request1 sources
- Heat index 596RisingThinking Machines Lab Open-Sources Inkling 2 Mini with 32GB Model Card Recommendation10 sources
- Heat index 78LookAtMyQuant open-sources compiler that quantizes Qwen3.5-4B to 1.23 GB1 sources
- Heat index 54MiniMax open-sources H3 Image for both image generation and editing1 sources
- Heat index 38Together AI launches Qwen3 TTS with 9 voices across 10 languages1 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

