flâneur

Claude Fable 5 & Claude Mythos 5 — AI System Cards

malob.github.io · 79,982 words · saved by 1 readers

System card for Claude Mythos 5 and Claude Fable 5, two configurations of a new large language model from Anthropic. Covers RSP evaluations, cyber capabilities, safeguards and harmlessness, agentic safety, alignment assessment, model welfare, and capabilities.

Changelog June 11, 2026 Noted CAIS’s contribution to the VCT CB-1 evaluation in Section 2.2.4.1. Corrected a minor error in our description of the results of Figure 7.2.3.A. Corrected “Mythos” to “Fable” on figures in Sections 8.4, 8.7, and 8.17.9. Corrected the effort level at which Fable 5 performed best on Vending-Bench from “max effort” to “high effort” in Section 8.17.6. Standardized spelling to American English (e.g. “behaviour” to “behavior”). Minor formatting corrections and standardization. In a previous version of the system card, we described the initial version of our frontier LLM

saved by

related reading