Anthropic has introduced Claude Opus 5, a new artificial intelligence model designed to deliver near-flagship performance at a significantly lower cost, as competition intensifies among leading AI developers and businesses increasingly demand more affordable AI services.
The San Francisco-based company says its latest model offers performance approaching its most powerful commercial system while requiring fewer computing resources, helping reduce operating costs for enterprises deploying large-scale AI applications.
The launch underscores a growing shift in the AI industry, where lowering inference costs is becoming just as important as improving model capabilities.
High-End AI at Half the Cost
According to Anthropic, Claude Opus 5 achieves performance close to its premium Fable 5 model across a wide range of coding, reasoning and knowledge-intensive tasks while costing only half as much.
The company has priced Opus 5 at $5 per million input tokens and $25 per million output tokens, making advanced AI capabilities more accessible to developers and enterprise customers.
Most commercial AI platforms charge customers based on the number of tokens processed, meaning lower token pricing can significantly reduce costs for businesses running large-scale AI workloads.
Anthropic believes this combination of competitive performance and lower pricing will appeal to organisations looking to expand AI adoption without dramatically increasing infrastructure expenses.
Competition in AI Is Shifting Toward Cost Efficiency
Claude Opus 5 enters an increasingly competitive market dominated by companies such as OpenAI and Google, whose ChatGPT and Gemini models continue to compete aggressively for enterprise customers.
While early AI competition focused primarily on building the most capable models, the market is now placing greater emphasis on operational efficiency and affordability.
Businesses deploying AI at scale often process billions of tokens each month, making model pricing a major factor in purchasing decisions.
As a result, AI companies are increasingly competing not only on benchmark performance but also on cost per query, response quality and infrastructure efficiency.
Investor Scrutiny Intensifies Ahead of Public Listings
Anthropic’s latest product launch comes at a critical time for the company.
Both Anthropic and OpenAI recently filed for public stock market listings, bringing increased attention from investors seeking evidence that leading AI companies can build sustainable, profitable businesses.
While demand for generative AI continues to grow rapidly, operating advanced models remains extremely expensive because of the enormous computing resources required for training and inference.
Reducing serving costs while maintaining competitive performance has therefore become one of the industry’s highest priorities.
Lower-cost models like Claude Opus 5 could help improve profit margins while expanding enterprise adoption.
Anthropic Continues Emphasising AI Safety
Alongside performance improvements, Anthropic says it has maintained a cautious approach to AI safety.
The company stated that Claude Opus 5 was intentionally not trained extensively on offensive cybersecurity tasks, reducing the likelihood that the model could be used to generate sophisticated cyberattacks.
Anthropic also says the model remains substantially less capable than its highly restricted Mythos 5 system in areas involving offensive cybersecurity and advanced biological research.
Mythos 5, available only to selected partners, has generated significant discussion because of its ability to identify software vulnerabilities and develop exploit strategies that could potentially be misused.
By limiting those capabilities in Opus 5, Anthropic aims to balance strong technical performance with lower operational risk.
New Model Reduces Safety Interventions
Interestingly, Anthropic says Claude Opus 5 performs nearly as well as Mythos 5 when identifying software vulnerabilities but is significantly less capable of converting those discoveries into practical cyberattack methods.
Because of this lower risk profile, the company expects its automated safety systems to intervene approximately 85% less frequently compared with Fable 5.
Fewer safety interruptions could improve the user experience for legitimate developers while maintaining safeguards against potentially harmful requests.
The approach reflects Anthropic’s broader strategy of tailoring safety measures to each model’s capabilities rather than applying identical restrictions across every system.
Export Controls Previously Disrupted Model Availability
Anthropic’s newest launch follows a period of regulatory uncertainty.
Earlier this year, both Fable 5 and Mythos 5 were temporarily removed from public availability after the U.S. Commerce Department introduced export controls affecting advanced AI technologies.
Those restrictions were later lifted, allowing Anthropic to restore access to the models.
The episode highlighted the increasingly strategic importance of frontier AI systems, as governments around the world continue evaluating how advanced artificial intelligence should be regulated and exported.
The AI Race Is Entering a New Phase
Claude Opus 5 reflects a broader evolution within the AI industry.
Instead of focusing exclusively on creating ever-larger and more powerful models, leading AI companies are increasingly investing in cost optimisation, efficiency and commercial scalability.
For enterprise customers, affordability has become almost as important as raw performance.
Businesses want AI systems capable of handling coding, research, analysis and customer service tasks without generating unsustainable operating costs.
Anthropic believes Opus 5 addresses that challenge by delivering near-premium performance at substantially lower prices while maintaining robust safety controls.
As competition between Anthropic, OpenAI, Google and other AI developers accelerates, the next phase of the AI race may be determined not only by which company builds the smartest model, but also by which can deliver advanced intelligence at the lowest cost and highest efficiency.
