Every 团队 7 人中 6 人给 Claude Fable 5.2 金奖：适合可放手的长任务
every · x · 2026-11-10
Every's team tested Anthropic's new Claude Fable 5.2 on coding, design and writing work.
- 6 of 7 testers gave it gold, their top rating for models that unlock new possibilities
- The enthusiasm centers on one use case: ambitious projects you can hand off and leave running
- Sole holdout Dan Shipper: "It makes things possible that weren't possible before, but it's narrow, and it's expensive"
Every is running a livestream, team discussion and a camp on configuring Fable 5.2 for long jobs; the full Vibe Check lands soon.
所属事件:Every 团队实测 Claude Fable 5.2：长任务可放手，档位并非越高越好(2 条相关)→
「模型」频道最新
- Reddit 用户抱怨免费版 Gemini 越来越差 — rupomthegreat · 2026-11-10
- Terminal-Bench-Science 最高分四个月从 30% 冲到 74.6%，硬科学任务告急 — geoffwolfe · 2026-11-10
- Fable 领跑 Slopalytics 榜单5分，Zoink 直呼 Dario 统治光锥 — beffjezos · 2026-11-10
- 创始人实测吐槽：新版更贵还更容易错，回退旧版当主力 — bindureddy · 2026-11-10
- 研究者辨析:Fable 5.2 疑为机器人数据训练的 VLM 加 agent 框架 — ZeeshanZiaML · 2026-11-10
- GPT-6 Astra 对决 Fable 5.2：像素画生成能力现场比拼 — Angaisb_ · 2026-11-10