DaisyRose-27B: 1,200-task RL took Qwen3.8 from 5.6% to 12% on Terminal-Bench 4.1
discuss
- Points193
- Comments63
discuss
Story details from Hacker News ↗ · linked article on radixark.com.
DaisyRose-27B: 1,200-task RL took Qwen3.8 from 5.6% to 12% on Terminal-Bench 4.1 was first seen on Hacker News · Front page on Nov 18, 2026. Since then it has appeared in 72 checks on 2 different days. Its best position was #4 of 30, reached on Nov 18, 22:13 UTC. It dropped off the list after Nov 19, 13:58 UTC.
Position on Hacker News · Front page at every check (lower is better).
Best rank per day (UTC).
Platform metric as reported by Hacker News.
A few top-level comments from Hacker News, shortened. Read the full thread on Hacker News.
Very excited to see more post-training happening on these open models. Looking forward to trying this one out once it makes its way onto OpenRouter. Qwen 27B really punches above its weight. 27B is small enough to fit on a Mac Mini / Macbook Pro which is very exciting. Starting to see more promising examples of…
The heavy reliance on eval datasets showcased here and also evidenced by the frontier lab's high payments for such data cast doubt on the claims that AI will become self-improving soon
Since some talented people with more modest means were able to do such a job, I wonder why Alibaba didn't do something like this when they trained the model. They left a lot of performance on the table. Which makes me wonder if such kind of optimizations are possible for other models like Deepseek v4.1 Flash 1017.
Archived daily, weekly and monthly lists that include this entry.
Not at the moment. It was last on the list on Nov 19, 13:58 UTC, after 2 days of trending.
#4 on Hacker News · Front page, on Nov 18, 22:13 UTC. Trendoday checks this list every 15 minutes.
It was first seen on Nov 18, 2026 and has been on the list on 2 days (72 checks) since.
193 points and 63 comments; submitted by de_sty on Nov 18, 2026.