Anthropic’s Claude Fable 5 and Mythos 5 LLMs: Launch, Access, Pricing, Safety
Generated: 2026-06-11 · API: Gemini 2.5 Flash · Modes: Summary
Anthropic’s Claude Fable 5 and Mythos 5 LLMs: Launch, Access, Pricing, Safety
Clip title: Mythos 5 & Fable 5 Launched Author / channel: Sam Witteveen URL: https://www.youtube.com/watch?v=PkYwtQcGIsU
Summary
Anthropic has officially launched two new large language models: Claude Fable 5 and Claude Mythos 5. While Claude Fable 5 is now generally available to a broader audience, Claude Mythos 5, which represents a higher capability level, is currently restricted to select partners involved in “Project Glasswing” (cyber defenders and critical software infrastructure providers) and soon to specific biology researchers, with additional safeguards. These models are touted as “Mythos-class” and reportedly surpass previous models like Claude Opus 4.8 in benchmarks across diverse areas including software engineering, knowledge work, vision, and scientific research.
The pricing strategy for Fable 5 and Mythos 5 has been adjusted to be more accessible, with costs set at 50 per million output tokens. This represents a significant reduction, being less than half the price of the previous Claude Mythos Preview. Regarding availability for existing users, Fable 5 is temporarily included without extra cost for those on Pro, Max, Team, and seat-based Enterprise plans until June 22nd. After this date, usage credits will be required, though Anthropic aims to eventually restore Fable 5 as a standard part of subscription plans as capacity allows. The video suggests that recent infrastructure deals might be enabling Anthropic to offer these models at a more competitive price point.
A key aspect of Fable 5 is the implementation of new “safety classifiers.” These AI systems are designed to detect potential misuse, including “jailbreak” attempts, and to prevent the model from responding to requests related to cybersecurity, biology, chemistry, or distillation that could pose significant risks. If a request is flagged, the model automatically switches to a less capable model, Claude Opus 4.8, to continue the conversation. The presenter demonstrates this conservative approach by attempting to ask about Ebola outbreaks and the World Cup, which immediately triggers the safety classifier, switching to Opus 4.8. While intended to prevent harm, this aggressive filtering might limit the model’s utility for seemingly benign, though sensitive, inquiries.
Benchmarking data presented in the video indicates that Claude Mythos 5/Fable 5 achieves notably higher scores in areas like agentic coding (80.3% on SWE-Bench Pro compared to Opus 4.8’s 69.2%) and legal agent benchmarks (13.3% versus Opus 4.8’s 10.4%). However, improvements in other categories, such as tool use and computer use, are less pronounced. Additionally, Anthropic has introduced a 30-day data retention policy for all traffic on Mythos-class models to defend against complex attacks, assuring users this data will not be used for training new models for non-safety-related purposes. Overall, while the new models exhibit superior capabilities, their practical application for general users may be significantly influenced by these stringent safety measures and the evolving subscription model.
Video Description & Links
Description
This video, I cover the latest release from Anthropic, which is the new Fable 5 model, which is a Mythos Class model. We have a look at the first initial benchmarks, the pricing, and also some of the security restrictions on this model.
Blog: https://www.anthropic.com/news/claude-fable-5-mythos-5 warnings: https://support.claude.com/en/articles/15363606-why-claude-switched-models-in-your-conversation-with-fable-5
Twitter: https://x.com/Sam_Witteveen
🕵️ Interested in building LLM Agents? Fill out the form below Building LLM Agents Form: https://drp.li/dIMes
👨💻Github: https://github.com/samwit/llm-tutorials
⏱️Time Stamps: 00:00 Mythos Rebrand Reveal 01:26 Access Tiers Explained 02:04 Pricing Reality Check 02:50 Benchmark Highlights 05:05 Safety Limits Preview 05:40 Data Retention Shift 07:08 Subscription Changes Ahead 08:38 Fable Demo 11:32 Biology Trigger Test 12:50 Wrap Up Next Steps
Tags
mythos, fable, fable5