Anthropic on Friday unveiled Claude Opus 5, its newest flagship artificial intelligence model, positioning it as its strongest balance of performance and affordability as competition among leading AI developers increasingly shifts from raw capability to commercial value and cost efficiency.
The San Francisco-based AI startup said Opus 5 outperforms its previous flagship, Claude Fable 5, across key benchmarks for software engineering, coding and knowledge work while cutting usage costs by 50%. Anthropic also said the model is designed for everyday enterprise workloads rather than niche or experimental applications.
Vendors in the AI industry are now under pressure to justify the enormous investments being poured into model development and AI infrastructure. As enterprises become more selective about AI spending, model providers are competing not only on benchmark performance but also on cost, efficiency, and measurable business outcomes.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Claude Opus 5 will cost $5 per million input tokens and $25 per million output tokens, compared with significantly higher pricing for Fable 5. Anthropic said the lower pricing does not come at the expense of capability, describing Opus 5 as its best-performing and most cost-effective model across multiple industry evaluations.
The company added that while Opus 5 delivers stronger performance in coding and knowledge-intensive tasks, it is not its most capable model for high-risk dual-use applications, such as offensive cybersecurity research. That distinction remains with Claude Mythos 5, Anthropic’s specialized cybersecurity-focused model.
The announcement comes at a time when the economics of artificial intelligence are becoming as important as technical leadership. Companies are deploying AI at scale but are increasingly demanding lower inference costs, predictable pricing and stronger returns on investment after years of heavy infrastructure spending.
“Enterprises, in our feedback and with our customer base, are looking for value,” Dianne Penn, Anthropic’s Head of Product Management for Research, told CNBC.
“If it’s a cheaper model or a cheaper offering, but it’s not accomplishing a similar level of quality, it’s actually not useful.”
The pricing move also underlines the mounting competitive pricing pressure across the AI industry.
Anthropic is competing against OpenAI, Google, Microsoft and Amazon, while Chinese developers including Moonshot AI, Alibaba, Z.ai and MiniMax have introduced capable open-weight models at substantially lower operating costs. Those releases have intensified pricing competition and challenged assumptions that frontier AI models must remain expensive to operate.
The latest model also arrives as investors scrutinize AI companies’ spending more closely. Industry-wide capital expenditures on AI infrastructure continue to surge into the hundreds of billions of dollars, prompting customers to seek models that can deliver comparable performance with lower operating costs.
Anthropic’s strategy suggests the company is attempting to expand beyond customers willing to pay premium prices for frontier capabilities by offering a model that balances performance with commercial practicality.
The release follows a turbulent few months for the company.
In April, Anthropic introduced Claude Mythos Preview, a cybersecurity-focused model that demonstrated advanced vulnerability discovery and exploitation capabilities during controlled testing. Anthropic later launched Mythos 5 alongside Claude Fable 5 in June, describing Fable 5 as its most capable general-purpose model to date.
Shortly after those releases, the U.S. government temporarily suspended access to both models under national security-related export controls before lifting the restrictions roughly two weeks later following discussions between Anthropic and government agencies.
Penn said Anthropic continues to work closely with U.S. authorities during model evaluation and deployment.
According to the company, Opus 5 remains less capable than Mythos 5 in sensitive areas such as offensive cybersecurity and biological research, reflecting Anthropic’s continued separation between its commercial AI offerings and its highest-capability research models.
However, the release of Opus 5 is seen as an indication that leading developers are beginning to emphasize cost efficiency, inference economics, and practical enterprise deployment.