Anthropic’s Claude Fable 5 is a version of Mythos the public can access today

Anthropic has officially released Claude Fable 5, marking the first time the general public can access its advanced AI technology. This model, a variant of the previously private Mythos, is designed with strict safety protocols to prevent misuse. Fable 5 is touted for its exceptional capabilities in software engineering, knowledge work, and vision. However, in critical fields such as cybersecurity, biology, and chemistry, the model is programmed to revert to its predecessor, Claude Opus 4.8, to maintain safety standards. Initially, Mythos was restricted to a select few partners due to security concerns, but access has since broadened to hundreds of organizations across 15 countries, particularly those managing essential infrastructure. The new model is accessible via Anthropic’s Claude API and is included in various subscription plans with a phased rollout. From now until June 22, users on Pro, Max, Team, and seat-based Enterprise plans can utilize Fable 5 at no extra cost. However, starting June 23, it will require usage credits, with plans to reintroduce it as a standard subscription feature in the future. In parallel, Anthropic is releasing an upgraded version of Mythos, known as Mythos 5, to organizations already cleared for the advanced model. This launch coincides with Anthropic’s preparations to enter the public markets, alongside competitors like OpenAI and SpaceX. Anthropic has expressed concerns over the rapid evolution of AI, particularly regarding recursive self-improvement, which could lead to autonomous advancements without human oversight. To mitigate risks, the company conducted extensive testing, including an external bug bounty program, which did not yield any universal jailbreaks during over 1,000 hours of examination. Nonetheless, Anthropic acknowledges the potential for new types of attacks. With the introduction of Fable 5 and Mythos 5, the company will implement a 30-day data retention policy on all traffic, a shift from previous zero-retention agreements. This data will not be used for training but will aid in safeguarding against future threats and refining the model’s performance. Early reports suggest that Fable 5 effectively handles a majority of queries independently, with over 95% of sessions relying solely on its own responses. Analytics firm Hex praised Fable for achieving a 90% score on its complex task benchmarks, highlighting its strong judgment and nuanced understanding. Additionally, developers like Base44 and Genspark reported that Fable outperforms other models in tasks such as UI design and application development. The pricing for Fable 5 and Mythos 5 stands at $10 per million input tokens and $50 per million output tokens, which is double the cost of Opus 4.8. This pricing may deter some enterprises that are already scrutinizing their AI expenditures. Despite the costs, demand for Fable 5 is anticipated to be high. Companies like Rakuten believe the investment is justified, citing Fable's ability to enhance operational autonomy through its advanced reasoning capabilities.