BREAKING: Nous dropped Arete-27B.
This is a dev post classified by AGK Intelligence as AI dev tools (a launch), kept by the Dev Radar because it carries real work, not commentary.
BREAKING: Nous dropped Arete-27B. Base: Qwen3.8-27B. Post-train: ~53M tokens. Result: answers that adopt an official party line fell from 22% to 1% on their 803-prompt eval. Hermes Bench stayed flat (65.60% vs 65.77%). Open weights are not only a price decision. They are a control surface. If your agents will do fact-finding or judgment work, the values layer is late, thin, and editable. Fifty million carefully chosen tokens moved it. Capability underneath did not move. Weights, eval, and training pairs are public. Run the eval on whatever you ship. What values layer are you actually depl
Posted by Guillermo Flor (14.1k followers) 2 h ago · 0 likes · 172 views · view the original post on X. Kept by the Dev Radar as AI dev tools.
More dev work like this
- Stanford researchers gave an 8B language model 108× fewer document tokens. It became… — @techNmak
- 刷到一条好看的 MG 动画，想让 AI 照着这个风格做一条，提示词改了好几版，出来的还是差点味道。 — @GitHub_Daily
- Working on @UnderdogAI’s Woof on my Mac. — @plotarmordev
- 😨惊了！Kimi K3.1 Pro + Cerebras 直接把浏览器代理玩成开挂模式！ — @NFT_Chen
- 土豆姐最近有一场约 1 小时的工作坊。讲的是怎么一步步做到放心让 agent 自己写代码，甚至自己合 PR：一个月能合 1000 个 PR，早上起来就有 20 个已经合进主干。 — @mylifcc
- Your Claude window runs out mid-task. Does your agent just stop? 😩 — @BkashJosi
- Your AI scripts need more than a working demo. — @DanKornas
- Y Combinator 开源了内部使用的 AI Agent 框架 QM（Quartermaster），重点解决多人共用 Agent 时的权限和数据隔离问题。 — @bkdgiffug
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.8k posts from 5k X accounts over the last 21 days, 9.6k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-07 13:43 UTC. Full method.