Back to News

Anthropic launches Claude Fable 5.2 and Mythos 5.2 with faster speed, lower prices

856 points · 975 comments#anthropic#claude#model-release#ai-pricing

Anthropic released two new flagship models, Claude Fable 5.2 and Claude Mythos 5.2, claiming improvements in coding, reasoning, vision, and communication. The models generate outputs up to 30% faster than the 5.1 versions, with API pricing cut 20% to $8 per million input tokens and $40 per million output tokens; an optional fast mode doubles the price for 2.5x speed. During prerelease testing, the models formalized a proof of Fermat's Last Theorem in 30 hours, solved two open black hole physics problems, and achieved a 37.3-point jump on the RoboDojo Sim benchmark.

Coverage timeline

  1. The Verge AIJay Peters

    Anthropic is announcing two new models today: Claude Fable 5.2 and Claude Mythos 5.2, which it says are designed to take on "daunting" technical challenges. The company claims the new models show improvements in coding, reasoning, and vision, and can generate outputs faster than the 5.1 versions of the Fable and Mythos models. Claude Fable 5.2 is good at challenging software engineering problems, Anthropic says, and can orchestrate agent teams and tackle long-running projects. The new model has also solved some complex scientific problems, including two open theoretical physics problems. During prerelease testing, the two models have "help … Read the full story at The Verge.

  2. Techmeme

    Anthropic : Anthropic releases Mythos 5.2 and Fable 5.2, saying they are “best for the most ambitious, open-ended technical work” — We've upgraded Claude Fable and Claude Mythos—models that are designed to tackle the most daunting technical challenges. — Claude Fable 5.2 …

  3. Techmeme

    Jay Peters / The Verge : Anthropic says during prerelease testing Mythos 5.2 and Fable 5.2 formalized the proof of Fermat's Last Theorem and solved two open black hole physics problems — Claude Fable 5.2 is designed to take on big technical challenges, and it's available today.

  4. TechCrunch AIRussell Brandom

    Anthropic says the new models "show improvements in areas like coding, reasoning, and vision," as well as in communicating more clearly with users.

  5. Anthropic News

    Our most capable models for daunting technical work, Claude Fable 5.2 and Claude Mythos 5.2 set a new bar in coding, finance, and the sciences.

  6. Techmeme

    Frederic Lardinois / The New Stack : Anthropic prices Fable 5.2 and Mythos 5.2 at $8/1M input and $40/1M output tokens, down 20% from 5.1; fast mode costs $16/1M input and $80/1M output tokens — On Monday, Anthropic launched Claude Fable 5.2 and Mythos 5.2, the latest versions of its duo of flagship models.

  7. 量子位量子位

    Fable 5.2深夜掀桌!从0造操作系统,性价比反超Opus 5.5 梦晨 2026-11-10 06:08:18 来源: 量子位 在训练阶段主动锁死了新模型RSI能力 梦晨 发自 凹非寺 量子位 | 公众号 QbitAI 为冲刺IPO,A\掀桌了: Claude Fable 5.2 与 Mythos 5.2直接发布 。 有多强呢?从一个空文件夹开始,1000多个Agent昼夜不停连轴转8天,自主编写操作系统。 没有人类插手写代码,一套带内核、磁盘与网卡驱动、文件系统、窗口界面甚至安装程序的完整操作系统,直接在笔记本电脑上跑通开机。 甚至为了防止AI“自我迭代失控”,Anthropic在训练阶段主动锁死了新模型的RSI能力,把自我改进的能力限制在Opus 5.5的水平。 跑分你就看吧,仅有一项不是第一。 相比上一代5.1,新模型生成速度拉升30%,API定价直接下调20%至 输入$8/百万token、输出$40/百万token ; 如果愿意掏双倍价钱 ($16/$80) ,还能开启2.5倍速的Fast模式。 内测用户中,@mattshumer_ 制作了一个全场景可破坏的3D体素世界,物理引擎非常强大。 @kevin_t_ngo只做了一个浏览器运行的8个梦幻划船旅行场景,审美也非常在线。 Anthropic官方更是卷向了千级Agent集群协同与物理世界跨界工程。 1000个Agent自己拉群干活,从写系统到造硬件全包圆 目前Anthropic自身核心代码库中,超过90%的合并提交已经由Claude包办。 在内部测试中,Fable 5.2组成的Agent小队直接接管了一项核心生产服务,将其从Python完整迁移至Rust,服务启动速度暴涨26倍; 量化巨头Jump Trading将其接入内部交易研究框架后,Fable 5.2单枪匹马包揽了规划、协调、写代码和提交任务四个角色,免去了人工重新分配多模型权重的麻烦。 这种长周期、多角色的自主工程能力,在数学、物理、生物和硬件四个硬核领域给出了更直观的案例: 第一件事,30小时通关费马大定理形式化证明。 将数学证明翻译成计算机可逐行检验的Lean语言,是极其枯燥的苦力活。此前Anthropic用内部研究版模型配合人工指导和定制框架,花了整整11天。 这次,官方只给了极简提示词和中间定理列表,没给任何数学参考资料,直接放出10

  8. 机器之心机器之心

    终于,在新一轮旗舰模型谁先发的竞赛中,Anthropic 率先出招了。 就在今天凌晨,Anthropic 一口气发布两款顶级模型,Fable 5.2、Mythos 5.2 登场。 官方宣称,「两款模型瞄准高难度技术挑战,能够参与多个领域的大型复杂项目,并在编程、推理和视觉等方面实现进一步提升。」 另外,Anthropic 也开始玩起了 OpenAI Tibo 熟悉的「Reset」福利了。 Claude Code 之父还宣布,「我们希望为那些陪伴我们最久的用户提供更多福利。因此,在 Fable 5.2 中,订阅时间越长,用户获得的 Fable resets(重置次数)就越多。」 根据介绍,Claude Fable 5.2 和 Claude Mythos 5.2 在正式发布前的测试中,帮助 Anthropic 的工程师完成了此前难以实现的软件迁移项目,从零设计出三款能够正常运行的电子设备,解决了两个黑洞物理学领域尚未解决的问题,在机器人评测中取得了重大进展,还仅用 30 小时就完成了费马大定理的形式化证明。此外,它们还有许多其他应用场景。 Claude Fable 和 Claude Mythos 使用相同的底层模型。Fable 配备了严格的网络安全和生物安全防护机制,可以有效防止模型被滥用,因此能够向所有用户开放。对于希望将模型用于网络安全或生物医学研究等有益领域的机构,可以通过 Anthropic 近期扩大的生命科学计划和网络安全验证计划申请 Mythos 的使用权限。 两款模型的生成速度最高比 Claude Fable 5.1 和 Claude Mythos 5.1 快 30%,价格也有所降低: 每百万输入 Token 收费 8 美元,每百万输出 Token 收费 40 美元 。 Fable 5.2 和 Mythos 5.2 还提供可选的快速模式(Fast Mode),最高可以将 Token 生成速度提升至原来的 2.5 倍。快速模式的价格为每百万输入 Token 16 美元,每百万输出 Token 80 美元。 Claude Fable 5.2 和 Mythos 5.2 的评测表现 基准测试成绩 下方的基准测试结果显示,在 Anthropic 评估的几乎所有能力上,Claude Fable 5.2 和 Mythos 5.2 都比 Opus 5.5 更胜一筹。 从横

  9. Simon Willison

    Claude Fable 5.2 responds closely to what the prompt says about who owns the code it works on. In Anthropic’s testing on small coding tasks, when a prompt said nothing about ownership or presented the code base as someone else’s, the model tended to add to the existing structure, even when that repeated existing code. [...] If you want the model to improve the structure as it works, say that the code is its own: This codebase is yours: what it's for and how it's built are your call, and you'll keep building on it. In Anthropic’s testing on small coding tasks, the model merged repeated code in about half of the tries with this sentence, and in under a tenth of the tries without it. — Prompting Claude Fable 5.2 , Anthropic documentation Tags: prompt-engineering , anthropic , generative-ai , ai , llms , claude-mythos-fable

  10. Techmeme

    Ethan Mollick / One Useful Thing : Anthropic's Fable 5.2 represents “Jagged AGI” that matches human expert levels on select tasks like game design, but lacks self-directed judgment and taste — General, often expert, and sometimes jagged … I have had the chance to use the new Claude Fable 5.2 quite a bit …

  11. Techmeme

    Maximilian Schreiner / The Decoder : Scale Labs and CAIS' Remote Labor Index: Fable 5.2 completes 41.3% of freelance projects at professional quality, above Opus 5.5's 21.3% and GPT-6 Astra's 20.8% — - Fable 5.2 breaks through in 3D models, audio, and architecture … The test measures how often AI agents complete real freelance jobs …

  12. 机器之心机器之心

    今天凌晨,Anthropic 一口气发布两款顶级模型 Claude Fable 5.2 和 Mythos 5.2。新模型在科学研究方面展示了多项成果,包括通过约 1000 个 AI 智能体协作,仅用 30 小时完成费马大定理证明的形式化。 在此次更新中,机器人控制能力也备受关注。在官方的发布文章中,科学与工程部分专门设置了「硬件与机器人」小节。 官方演示视频中,Claude Fable 5.2 控制机器人完成了操作益智方块、拧开瓶盖、绘制机器人图案等任务。 根据 Anthropic 公布的数据,Claude Fable 5.2 在 RoboDojo Sim 上的得分达到 64.9%,较上一代 Fable 5.1 的 27.6% 提升 37.3 个百分点,也超过了 Claude Opus 5.5(39.4%)和 GPT-6 Astra(29.8%)。 Anthropic 还表示,Fable 5.2 在此次评测中超过了多数专门面向机器人控制开发的 AI 系统。 Anthropic 产品经理 Alex Albert 直言机器人的「苦涩的教训」已经来了。 与此同时,曾参与特斯拉 Optimus 人形机器人研发的研究员 Bingjie Tang 也已加入 Anthropic。她此前主要从事机器人灵巧操作、策略学习及真实环境部署,其研究背景与 Anthropic 近期展示的机器人能力形成了值得关注的交集。 Anthropic 正在拓展 Claude 的机器人控制能力 实际上,Anthropic 今年已经多次公开机器人相关研究。 7 月,Anthropic 发布《Claude plays robotics》,测试 Claude 在仿真机器人、机械臂以及真实四足机器人上的表现。 研究人员尝试了多种控制方式,包括生成底层控制指令、编写控制程序,以及调用预训练机器人策略。 实验发现,Claude 已经具备一定的机器人环境理解和操作规划能力,具体表现则受到控制接口、动作精度和推理延迟等因素影响。 到了 10 月,Robocurve 等研究机构发表论文《The Bitter Lesson Comes to Robotics》,进一步评估前沿通用多模态模型控制真实机械臂的能力。 该研究发现,通用模型已经能够完成多种实体机器人操作任务,相关能力在数月内持续改善。与此同时,操作精度、执行速度和推

  13. 机器之心机器之心

    编辑|山辉 继 OpenAI 之后,Anthropic 也开始抢机器人公司的风头了。 今天,Anthropic 发布了新模型 Claude Fable 5.2。除了常规的编程、推理能力升级,这次还有一个挺意外的看点:Claude 开始玩魔方、拧瓶盖,甚至能操控机器人画画了。 前段时间,OpenAI 的 Astra 就已经在机器人操作任务上秀过一把。如今,Anthropic 也跟了进来。 过去,这些能力还是具身智能公司费尽心思训练专用模型才能实现的。现在,通用基础模型也开始进入这块地盘。 这不免让人担心,机器人公司辛辛苦苦训练出来的模型,会不会有一天也被基础模型「吞掉」? 这个问题已经吵了一段时间,目前还没有定论。但无论最后控制机器人的是通用大模型、VLA 还是世界模型,大家都逃不过一个基础问题:它们对物理世界的经验从哪里来? 物理经验可以单纯从通用基础模型所用的常规数据中涌现吗?至少前段时间针对 Astra 的一项研究表明:事情没这么简单,OpenAI 在训练过程中可能使用了专门的机器人数据,GPT-6.1 Sol 和 Gemini Argon 也接受了类似训练(参见 《GPT-6 Astra 的机器人能力被扒:可能接受过专门训练》 )。 也就是说,这轮通用模型在机器人任务上的高光表现,未必是智能凭空涌现,更可能是数据在背后托底。 即使是通用大模型,也需要通过训练数据补足对真实世界的理解。随着模型能力不断提升,具身智能能走多远,越来越取决于背后数据基础设施的支撑。 特别是涉及真实物理接触的数据信息,恰恰是机器人学习精细操作的关键。 触觉数据也因此受到越来越多前沿机器人研究团队的关注。比如 Meta FAIR 推出的 Digit 360 触觉传感器和 SparAsh 触觉模型,就在尝试让机器获取并理解更丰富的接触信息。这也让触觉数据变得越来越重要。 这件事,国内确实有人在认真做。 自成立以来,简智一直站在这条产业链的更上游:建设数据基础设施。从第一视角的头戴设备,到戴在手上的数据手套,他们想解决的始终是同一个问题 —— 怎么把人在真实世界里的操作,尽可能完整、准确地变成模型能学的数据。 近日,他们更新了这条产品线,推出了新品:数据手套 Dex 2.0,和第一视角头戴设备 DAS Ego 2.0。 围绕第一视角视觉、腕部近场视觉、手部运动与触觉,构建多层次的 Human

Related stories

  • Leaked Azure entries hint at Anthropic's Claude Fable 5.5 and next-gen Opus

    Screenshots from Azure's model catalog and user tests suggest Anthropic is preparing two new Claude models: Fable 5.5 and a next-generation Opus, possibly launching as soon as next week. The Fable 5.5 entry briefly appeared in Azure's catalog, and benchmark tests show a model codenamed Obsidian 3 scoring slightly above the current Opus 5.5. The leaks come amid Anthropic's upcoming IPO.

    1 reports · 1 sources
  • Anthropic: Claude flags 493 security issues across 25 open-source projects

    Anthropic reported that Claude identified 493 security issues across 25 open-source projects, covering over 5 million lines of C, C++, Java, PHP, and Python code, including one vulnerability that went unnoticed for 21 years. The lab also announced a partnership with Theorem to use Claude for formally proving security properties at the assembly level across more than 100 real-world open-source libraries, extending formal verification to larger codebases.

    4 reports · 3 sources460 points · 310 comments
  • Anthropic files S-1 for Nasdaq IPO under ticker ANT

    Anthropic, PBC filed a Form S-1 with the SEC for an initial public offering on the Nasdaq under the symbol ANT, with the number of Class A shares and price range yet to be determined. The 335-page prospectus reveals strong revenue growth, mounting losses, an unusual governance structure for founder control, and over $800 billion in long-term compute commitments.

    34 reports · 8 sources1703 points · 1674 comments
  • METR: Anthropic models improving twice as fast, AI use drives ~80% of acceleration

    A METR report, based on an 11-week internal assessment at Anthropic, finds model capability progress roughly doubled since October 2025, with AI-assisted research driving about 80% of the acceleration. The report also reveals three unreleased internal models (Model 2, 3, 4), with the strongest exceeding Opus 5.5 by about three months of progress. Anthropic redacted much of the most informative evidence from the public version, citing confidentiality concerns.

    6 reports · 5 sources47 points · 6 comments

News updates every hour. Subscribe in your RSS reader to get new stories automatically.

Subscribe