Claude Fable 5 and Claude Mythos 5 are the flagship models of Anthropic's Claude 5 family, introduced in 2026 as a new "Mythos-class" tier positioned above the long-running Opus line of the Claude family. The two names refer to the same underlying model offered on two tracks: Fable is the generally available version and carries additional safety measures for dual-use capabilities, while Mythos is offered without those measures only to approved organizations, such as vetted cyberdefenders and life scientists.
The two-track model
The Fable/Mythos split formalized an access pattern for frontier models whose strongest capabilities are dual-use: instead of withholding the model entirely or shipping it unrestricted, Anthropic gates the unmitigated variant behind organizational vetting while the safeguarded variant ships broadly. The approach is grounded in Anthropic's responsible scaling policy and was echoed days later by OpenAI, whose GPT-6 Astra launched with its most advanced cybersecurity capabilities limited to a vetted program.
Fable 5.1
On September 1, 2026, Anthropic released Claude Fable 5.1 and Mythos 5.1, describing them as its most advanced models for coding and knowledge work, delivering stronger results than Fable 5 while reducing costs for many workloads. Mythos 5.1 access remained limited to a set of vetted US organizations. The release landed at the start of a dense week for frontier launches, alongside Google's Gemini 3.8 Flash (September 2) and OpenAI's GPT-6 Astra (September 3).
Reception and benchmarks
On community leaderboards tracked daily by LMArena, Claude Fable 5 established itself at the top of the web-development arena and among the leaders in expert, math and coding text categories, competing with Anthropic's own Opus 5 tier. The family is frequently discussed in AI safety debates as a live experiment in whether tiered access can reconcile open availability with dual-use risk, a question that also shapes AI governance discussions around capability thresholds.