Anthropic releases Claude Opus 5, targeting cost-effective enterprise AI workloads

Anthropic has officially released Claude Opus 5, a new large language model designed for long-running agents, complex coding tasks, and professional knowledge work. The release marks a significant step change for the Opus tier, with the company stating that the new model delivers performance metrics close to its flagship Claude Fable 5 at half the cost. Opus 5 is now the default model on the Claude Max subscription tier and serves as the strongest option for users on the Claude Pro plan.

The model is positioned as a substantial improvement over its predecessor, Opus 4.8, offering state-of-the-art performance on software engineering and scientific research evaluations. On benchmarks such as Frontier-Bench v0.1 and GDPval-AA, Opus 5 outperforms all other models, more than doubling Opus 4.8’s performance on software engineering tasks while reducing the cost per task. In scientific research, particularly in organic chemistry and protein-related tasks, the model shows marked improvements over previous iterations, scoring significantly higher on internal benchmarks for inferring molecular structures and predicting protein sequence variations.

Pricing for the new model remains consistent with the previous Opus tier, set at $5 per million input tokens and $25 per million output tokens. This rate positions Opus 5 at half the cost of the comparable Claude Fable 5 and slightly cheaper than OpenAI’s GPT-5.6. The model is available across all platforms, including the Claude API, where developers can access it in a Fast mode that runs approximately 2.5 times the default speed at twice the base price.

Anthropic has emphasised enhanced safety and alignment protocols in the Opus 5 release. The company describes the model as its most aligned to date, exhibiting lower rates of deceptive behaviour and reduced susceptibility to misuse compared to Opus 4.8, Sonnet 5, and Fable 5. Specific safeguards have been implemented for cybersecurity and biological research to prevent the generation of material cyber threats or reckless biological actions. While Opus 5 is capable of identifying cybersecurity vulnerabilities, it remains substantially behind Anthropic’s Mythos 5 model in exploit generation, with classifiers designed to block binary-based vulnerability scanning and penetration testing.

The launch coincides with broader initiatives from Anthropic, including the release of the Anthropic Economic Index connector for Claude and a contribution of an additional $20 million to Public First Action, bringing total support to $40 million. The company also shared its research agenda for the Anthropic Economic Futures Research Fund, aiming to explore real data about AI and work. Opus 5 is now available for immediate use, with fallback options to Opus 4.8 enabled for flagged requests on the API and consumer platforms.