Claude Opus 5 is substantially stronger than Claude Opus 4.8 across the board, with the largest gains in agentic coding, computer use, and long-horizon knowledge work. It sets a new state-of-the-art on several third-party benchmarks and is comparable to or ahead of Claude Fable 5 and Claude Mythos 5 on many evaluations.
- Opus 5 is pitched as being as good as or better than Fable 5 on practical tasks, while being faster and costing half the price.
- The model deliberately avoided training on cyber-related tasks, meaning it lacks the full capability for complex cyber offense found in Mythos-class models.
- On virology evaluations, Opus 5 looks marginally better than Mythos 5, as these tasks involve individual steps rather than requiring large-scale reasoning.
- ArtificialAnalysis rates Opus 5 at a new high of 61, and the Anthropic ECI point estimate is 162.1, placing it on the Mythos trend line.
The article suggests that staying Opus-sized and avoiding cyber training provides time for defenders to access superior tools with classifiers that trigger 85% less often than Fable's.