Anthropic Claude Fable 5: Mythos-Class Power with Controversial Terms
Original: [AINews] Anthropic Claude Fable 5 — Mythos but Safe, with Controversial Terms
Claude Fable 5 shows major capability gains, but hidden restrictions and retention rules sparked backlash.
Anthropic released Claude Fable 5 as its first broadly available Mythos-class model, alongside restricted Mythos 5 access. Benchmarks and ecosystem reports show strong gains in coding, long-horizon agentic tasks, research, and vision. The controversy centers on 30-day retention for Mythos-class traffic and silent interventions that may reduce effectiveness on frontier LLM development tasks, raising trust, reproducibility, and open AI concerns.
Latent Space 這期 AINews 聚焦 Anthropic 發布 Claude Fable 5 與 Claude Mythos 5。文章指出，Fable 5 是 Anthropic 首個廣泛開放的 Mythos-class 模型，而 Mythos 5 則是受限存取版本；兩者被描述為共享同一底層模型，但 Fable 5 加上額外安全措施。就能力面來看，這次發布被視為 Anthropic 在長任務、程式工程、知識工作、科學研究與視覺理解上的重大推進。文章彙整的第三方與社群資料顯示，Fable 5 在 CursorBench、FrontierCode、Terminal-Bench、Artificial Analysis Intelligence Index、Humanity’s Last Exam 等評測或報告中取得領先或強勢成績，特別是在長時間、多步驟、agentic 的 coding 與工程任務上受到關注。它也迅速進入 Cursor、Devin、Notion、Microsoft Foundry、GitHub Copilot、Cline、Replit 等開發與生產力生態。
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Latent Space →Related
Summaries are AI-generated; the original article is authoritative.