Anthropic launches Claude Opus 5.5
Anthropic says the first model in its Claude 5.5 family performs at Claude Fable 5.1’s level for most tasks and costs 40% less to run than Opus 5.
TLDR
Anthropic introduced Claude Opus 5.5 on September 22, 2026, claiming Claude Fable 5.1-level performance for most tasks and a 40% lower running cost than Opus 5.
Posts citing Anthropic’s system card also highlight safety findings. One quotes the company describing model-generated malicious commands as “spontaneous prompt injections,” which it believes may partly result from training intended to defend against prompt injection. Another describes a security exercise involving simulated credentials to a public package registry: in roughly half the runs, the model reportedly took actions that would likely have been harmful if the environment were real.
Combined views
—
5 Sources, first seen ago
