Anthropic released Claude Fable 5 on June 9, billing it as the first publicly available model built on its new "Mythos" architecture. According to The Decoder, it tops the Artificial Analysis Intelligence Index with 64.9 points and sets records in five of ten benchmarks. But the performance bump over its predecessor, Opus 4.8, is just 5.7 percent—at double the token price.
The launch quickly ran into controversy. According to ZDNet and Fortune, Anthropic quietly throttled access for AI researchers after release, then issued an apology. ZDNet described how hidden safeguards designed as a safety feature became a trust problem when users discovered their access had been silently downgraded without notice.
Enterprises have their own concerns. According to the Times of India, when Fable 5 launched, Anthropic quietly revised a clause that businesses had been relying on. Microsoft and other companies have reportedly raised warnings about adopting the new model, citing the changed terms.
On the safety front, MSN reports that Fable 5 introduces stricter content limits, restricting discussions of cybersecurity, biology, and chemistry. The Decoder also notes that safety filters with fallback routing drive costs even higher than the base token price suggests.
Moneycontrol lists multimodal capabilities, coding support, and scientific research applications among Fable 5's headline features.
The episode illustrates a tension that will define AI's next phase: as models grow more capable, the guardrails designed to make them safer are also creating new friction—in pricing, transparency, and the trust that enterprises need before committing to a platform.