Anthropic released Claude Opus 5, describing it as “a thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price.”
Opus 5 is positioned as a step-change improvement for the Opus tier and is now the default model on Claude Max and the strongest option on Claude Pro. It keeps the same pricing as its predecessor Opus 4.8 - $5 per million input tokens and $25 per million output tokens - while delivering substantially higher performance per dollar. Anthropic emphasizes that the model is designed for daily professional use, long-running agents, software engineering, and complex knowledge work rather than pure research frontier tasks.
The company reports clear gains in agentic behavior, verification, and iteration. On several benchmarks Opus 5 sets new state-of-the-art results or approaches Fable 5 performance at markedly lower cost. On Frontier-Bench it more than doubles Opus 4.8’s score and leads all models. On ARC-AGI-3 its result is roughly three times higher than the next-best model. On CursorBench 3.2 and OSWorld 2.0 it reaches near or better than Fable 5 levels at approximately half the cost per task or less. It also shows meaningful improvements over Opus 4.8 on life-sciences evaluations, including organic chemistry, structural biology, and bioinformatics.
A cost-efficient workhorse
A Fast mode runs about 2.5 times faster at twice the base price, and an adjustable effort setting lets users balance intelligence against cost and latency. Beta features include mid-conversation tool changes and automatic fallbacks on the API.
On safety, Anthropic states that Opus 5 is its most aligned model to date on automated behavioral audits, showing the lowest rates of deceptive or misaligned behavior among recent releases. It remains behind higher-tier models such as Mythos 5 on certain dual-use capabilities and retains targeted safeguards similar to those applied to Opus 4.8.
Overall, Claude Opus 5 is presented as a highly capable, cost-efficient workhorse for coding, agentic workflows, and professional tasks rather than a pure leap beyond Anthropic’s absolute top models.