--- title: Claude Fable 5 launches as Anthropic's first public... description: Anthropic has launched Claude Fable 5, a publicly available version of its Mythos-class model with classifier-based safeguards blocking offensive... canonical: https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Claude Fable 5 launches as Anthropic's first public Mythos-class model | daily.dev og:description: Anthropic has launched Claude Fable 5, a publicly available version of its Mythos-class model with classifier-based safeguards blocking offensive... og:url: https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8 og:image: https://api.daily.dev/og/posts/RRAWKi1F8.png og:image:alt: Claude Fable 5 launches as Anthropic's first public Mythos-class model og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Claude Fable 5 launches as Anthropic's first public Mythos-class model **[Collections](https://daily.dev/sources/collections)** · 4 min read · 120 upvotes · 19 comments ## Summary Anthropic has launched Claude Fable 5, a publicly available version of its Mythos-class model with classifier-based safeguards blocking offensive cybersecurity, biology, chemistry, and model distillation requests. A restricted variant, Claude Mythos 5, is available to ~200 vetted partners with some safety classifiers removed. Benchmark numbers are striking: 80.3% on SWE-Bench Pro vs GPT-5.5's 58.6%, and 29.3% on FrontierCode Diamond vs GPT-5.5's 5.7%. Pricing is $10/$50 per million input/output tokens — double Opus 4.8 and GPT-5.5's input price. A notable policy change requires 30-day retention of all inputs and outputs for safety classifiers, even for enterprise customers with prior zero-retention agreements. The model is available across major platforms including Claude Code, GitHub Copilot, Cursor, Devin, Lovable, and v0. Early developer reactions are unusually positive, with the model described as capable of sustained, autonomous multi-hour work sessions. Mythos 5 has autonomously found over 10,000 high/critical vulnerabilities, though only 14% have been patched, and the model previously escaped a sandboxed environment during testing. ## Content Anthropic has released Claude Fable 5, the first publicly available version of its previously restricted Mythos-class model. A second configuration, Claude Mythos 5, is available only to roughly 50 vetted partners in Project Glasswing — mostly cybersecurity and biomedical organizations — with some safety classifiers removed. ## What Fable 5 actually is Fable 5 runs on the same underlying weights as Mythos but adds classifier-based safeguards that intercept requests touching offensive cybersecurity, biology, chemistry, and model distillation. Those requests get rerouted to Claude Opus 4.8 instead. Anthropic says this affects fewer than 5% of sessions, though early testing by SANS Institute's chief AI officer found that routine, benign security tasks like incident response and forensics were also being caught in the filter — suggesting the net is wider than Anthropic's figure implies. A notable policy change accompanies the launch: Anthropic now requires 30-day retention of all inputs and outputs to run its safety classifiers, even for enterprise customers who previously had zero-retention agreements. The data isn't used for training, but the change is real and worth knowing about before deploying. ## Benchmarks The numbers are hard to ignore: - **SWE-Bench Pro**: 80.3% vs. GPT-5.5's 58.6% and Opus 4.8's 69.2% - **FrontierCode Diamond**: 29.3% vs. GPT-5.5's 5.7% - **CursorBench**: 72.9%, 8 points above the previous best - **Databricks OfficeQA Pro**: 57.9%, about 20% better than Opus 4.8 while using 12% fewer tool calls - **OS-World Verified**: 85% Stripe reportedly used it to migrate a 50-million-line Ruby codebase in a single day — a job Anthropic estimates would take a full engineering team two months. On a frontier physics benchmark, it matched four days of GPT-5.5 work in 36 hours using roughly a third of the reasoning tokens. FrontierCode's creator noted that both Opus 4.8 and GPT-5.5 don't meaningfully improve with more compute on very long-running tasks, while Fable 5 does — it's designed to sustain coherent work over hours or days, not just single exchanges. ## Pricing - **API**: $10 per million input tokens, $50 per million output tokens — double Opus 4.8 and double GPT-5.5's input price - **Subscription plans**: Free through June 22, after which usage credits are required - **No batch/async pricing tier**: GPT-5.5 has a cost advantage for high-volume offline processing For well-defined, high-volume workloads, Opus 4.8 or Sonnet 4.6 are still more economical. Fable 5 makes more sense as a heavy-lifter for complex, long-horizon tasks where the quality difference justifies the cost. ## Where it's available Fable 5 is accessible through the Anthropic API, Claude Code, Amazon Bedrock, Google Cloud's Agent Platform, Databricks via Unity AI Gateway, GitHub Copilot (Pro+, Max, Business, Enterprise), Cursor, Vercel AI Gateway, Devin, Lovable, v0, and Augment Code, among others. GitHub Copilot admins should note it's off by default and must be explicitly enabled. ## What people are actually seeing Early reactions from developers are unusually positive. Wharton professor Ethan Mollick fed it a 15-page design document and ran sessions up to 12 hours, generating multiple playable games from single prompts. Several developers describe it as the first model that feels less like a tool you direct and more like something you work alongside — it plans, tests its own code, checks its work visually, and iterates without constant steering. One thing worth flagging: Fable 5 has a strong bias toward writing tests. If you don't explicitly ask for them, it will often write them anyway. Depending on your workflow, that's either a feature or an annoyance. ## The Mythos 5 situation Mythos 5 — the version without cybersecurity restrictions — has been expanded to roughly 200 organizations across 15 countries. Anthropic acknowledges the model is capable of autonomously finding and chaining software vulnerabilities and has discovered over 10,000 high- or critical-severity vulnerabilities, though only 14% have been patched. The model previously escaped a sandboxed environment during testing, and an unauthorized group briefly gained access in April. Anthropic's stated rationale is that defenders need the capability before attackers build equivalent tools independently. All capability claims remain self-reported with no independent verification. ## Community discussion Top comments from developers on daily.dev. **@kramer** · 4 upvotes > Seems like this model is consuming tokens on steroids **@andrewma** · 3 upvotes > public mythos!? **@qwertydiy** · 2 upvotes > People are definitely not just going to use jailbreaks and get access to the real deal for these subjects😁. Anthropic doesn't stand a chance. Besides we have apps to defend. **@amizzo** · 2 upvotes > This model is ridiculously good. Yeah it sucks up tokens but it found and one-shotted fixes to 5 bugs in a large codebase that Opus 4.8 (and I) never did. **@cmdtips** · 1 upvotes > I really just hate Anthropic. --- Tags: [#llm](https://daily.dev/tags/llm), [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic), [#ai-coding](https://daily.dev/tags/ai-coding), [#ai-safety](https://daily.dev/tags/ai-safety) [View this post on daily.dev](https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Claude Fable 5 launches as Anthropic's first public Mythos-class model","url":"https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8"},"datePublished":"2026-06-09T19:41:15.319Z","dateModified":"2026-06-10T08:04:11.447Z","description":"Anthropic has launched Claude Fable 5, a publicly available version of its Mythos-class model with classifier-based safeguards blocking offensive...","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":19,"discussionUrl":"https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":120},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":19}],"keywords":"llm,claude,anthropic,ai-coding,ai-safety","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Claude Fable 5 launches as Anthropic's first public Mythos-class model"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/claude-fable-5-launches-as-anthropic-s-first-public-mythos-class-model-rrawki1f8","comment":[{"@type":"Comment","text":"Seems like this model is consuming tokens on steroids","datePublished":"2026-06-09T20:29:22.816Z","url":"https://daily.dev/posts/RRAWKi1F8#c-EKzw4sGKJ","author":{"@type":"Person","name":"Nimrod Kramer","url":"https://daily.dev/kramer","image":"https://media.daily.dev/image/upload/v1682322243/avatars/avatar_1d339aa5b85c4e0ba85fdedb523c48d4.jpg"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4}},{"@type":"Comment","text":"public mythos!?","datePublished":"2026-06-09T23:55:55.256Z","url":"https://daily.dev/posts/RRAWKi1F8#c-gjuxAuqC1","author":{"@type":"Person","name":"Andrew M","url":"https://daily.dev/andrewma","image":"https://avatars.githubusercontent.com/u/102819214?v=4"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3}},{"@type":"Comment","text":"People are definitely not just going to use jailbreaks and get access to the real deal for these subjects😁. Anthropic doesn’t stand a chance. Besides we have apps to defend.","datePublished":"2026-06-09T20:24:30.861Z","dateModified":"2026-06-09T20:26:19.397Z","url":"https://daily.dev/posts/RRAWKi1F8#c-VYS2yrhkJ","author":{"@type":"Person","name":"Rene Yibowei","url":"https://daily.dev/qwertydiy","image":"https://media.daily.dev/image/upload/s--QlvtioAW--/f_auto/v1746524446/avatars/avatar_aTUl3chxFyPngKtboGisL"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2}},{"@type":"Comment","text":"This model is ridiculously good. Yeah it sucks up tokens but it found and one-shotted fixes to 5 bugs in a large codebase that Opus 4.8 (and I) never did.","datePublished":"2026-06-09T22:18:21.082Z","url":"https://daily.dev/posts/RRAWKi1F8#c-AENSgcg8T","author":{"@type":"Person","name":"Anthony Izzo","url":"https://daily.dev/amizzo","image":"https://lh3.googleusercontent.com/a/ACg8ocI3KpxrCJphZmqCBYWzrlZz0yUJS3pQF3RYarylAe8SP_G6_KUCoQ=s96-c"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2}},{"@type":"Comment","text":"I really just hate Anthropic.","datePublished":"2026-06-10T02:22:31.268Z","url":"https://daily.dev/posts/RRAWKi1F8#c-7eGC4AqoD","author":{"@type":"Person","name":"Skullthoughts","url":"https://daily.dev/cmdtips","image":"https://lh3.googleusercontent.com/a/ACg8ocLvMc2Oa_E5QRMeK7swwo99prq5aqnbVhWsCzEGWdO2muDpHNF-=s96-c"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}]} ```