Anthropic's Claude Opus 5 Is Cheaper Than Fable 5, But It Has to Prove the Benchmarks

Anthropic released Claude Opus 5 this week, positioning it as a lower-cost professional model for coding, research and business automation while keeping API pricing unchanged at $5 per million input tokens and $25 per million output tokens. The company said in its announcement that Opus 5 is designed to deliver frontier-level performance for everyday professional workloads while balancing capability, latency and cost. Anthropic says Opus 5 outperforms both Opus 4.8 and, in some tests, Fable 5, which costs $10 per million input tokens and $50 per million output tokens. Early customer results cited by Anthropic include Harvey reporting a 26% reduction in token usage for legal research and Zapier saying Opus 5 completed an internal customer-retention workflow that previous models could not. As of publication, neither the benchmark scores nor the customer claims have been independently reproduced by outside evaluators. Anthropic acknowledged that Mythos 5 remains stronger on cybersecurity tasks and that Opus 5 produces a slightly higher factual hallucination rate than Opus 4.8. The company said in its Responsible Scaling Policy that transparent reporting of both model capabilities and known limitations is essential for helping developers make informed decisions. Opus 5 has a May 2026 knowledge cutoff, several months newer than both Opus 4.8 and Fable 5.