Anthropic has introduced Claude Opus 5, the latest version of its flagship Opus family of AI models, positioning it as a more affordable option for enterprises that need high-performance AI for coding, research and knowledge-intensive work. The release comes just weeks after the company unveiled its Claude 5 series, reflecting the rapid pace of competition among foundation model providers as they race to improve capability while lowering inference costs.
According to Anthropic, Claude Opus 5 delivers performance approaching that of its more powerful Claude Fable 5 model while costing roughly half as much. The company said the model has been designed for everyday enterprise workloads, with particular improvements in software engineering, scientific research, reasoning and long-running agentic tasks. It also retains the same pricing as the previous Opus 4.8 model at $5 per million input tokens and $25 per million output tokens.
The launch marks Anthropic's fourth major model release in less than two months, following the introductions of Mythos 5, Fable 5 and Sonnet 5. The rapid rollout reflects an industry-wide trend where AI companies are introducing increasingly specialised models tailored for different workloads instead of relying on a single general-purpose model.
Anthropic said Opus 5 demonstrates notable gains in agentic coding, knowledge work, visual understanding and multidisciplinary reasoning. The company also claims the model is stronger at verifying its own work and iterating until tasks are completed successfully, reducing the amount of user intervention required during complex workflows. Internal benchmark results shared by Anthropic indicate that Opus 5 outperforms its predecessor across several coding and reasoning evaluations and, in some cases, exceeds the performance of Fable 5 despite its lower operating cost.
A key addition is a new "Fast Mode," currently available as a research preview, which allows customers to trade higher inference costs for significantly faster responses. Anthropic has also introduced an adjustable "effort" setting that enables developers to control how much computational reasoning the model applies to a task, offering greater flexibility between speed, capability and cost.
Alongside performance improvements, Anthropic has placed renewed emphasis on AI safety. The company described Opus 5 as its most aligned Opus model to date, saying it includes stronger cybersecurity safeguards and is less susceptible to misuse than earlier versions. Anthropic said the model performs well for cybersecurity research while remaining below the capability threshold of its more restricted Mythos-class models that are intended for specialised organisations.
Claude Opus 5 is available immediately through Anthropic's paid Claude plans, the Claude API and cloud platforms including Amazon Bedrock. It becomes the default model for Claude Max subscribers and the highest-capability option available to Claude Pro users. Enterprise customers using Amazon Bedrock will also benefit from features such as zero data retention and enterprise-grade security controls.
The release comes as enterprise customers increasingly scrutinise AI operating costs while demanding stronger reasoning capabilities for production environments. Rather than focusing solely on benchmark performance, AI vendors are now competing on efficiency, governance and deployment flexibility. Anthropic's latest release reflects that shift, offering organisations a model intended to balance advanced capabilities with predictable pricing and enhanced security.