Anthropic launches Claude Opus 5, a cheaper AI model for coding, agents and enterprise workflows

Anthropic released Claude Opus 5 on July 24, 2026, positioning it as a more affordable AI model that retains most of the intelligence of its premium Claude Fable 5. Priced at $5 per million input tokens and $25 per million output tokens, the same as its predecessor Opus 4.8, Opus 5 is now the default on Claude Max and the most powerful option on Claude Pro. Rather than competing solely on peak intelligence, Anthropic focuses on delivering efficient, cost-effective AI for everyday enterprise tasks.

Benchmark results show Opus 5 excels in coding and knowledge work, setting new state-of-the-art marks on tests like Frontier-Bench and GDPval-AA. It scored 43.3% on Frontier-Bench v0.1, more than doubling Opus 4.8’s 18.7% and surpassing Fable 5’s 33.7%, while costing less per task. On ARC-AGI 3, it achieved three times the score of the next best model. However, Opus 5 trails competitors like Mythos 5 in cybersecurity and biology research and is less suited for long-duration, autonomous tasks where Fable 5 remains superior.

Anthropic emphasizes token efficiency as a key advantage, with customers reporting fewer tokens used and faster task completion without sacrificing accuracy. This efficiency is critical as enterprise AI spending grows and inference costs become a significant budget factor. Opus 5 also features an adjustable "effort" setting to balance intelligence, speed, and cost.

The model demonstrates advanced self-verification capabilities, iterating on tasks until successful outcomes are achieved. Examples include creating a computer vision pipeline to interpret images and identifying root causes of software bugs beyond existing community patches. This self-checking reduces the need for human review, lowering hidden costs in enterprise AI deployment.

Safety remains a priority, with Opus 5 showing the lowest rates of misaligned behavior among Anthropic’s models. The company intentionally limited training on cyber offensive tasks, resulting in strong vulnerability detection but limited exploit development. Requests flagged by safety classifiers fall back to less capable models to reduce risk. On scientific research benchmarks, Opus 5 outperforms previous models but is still behind Mythos 5 for complex, long-term projects.

Anthropic’s launch comes amid rapid commercial growth, with a reported $380 billion valuation and projected revenues rising from $1 billion in 2024 to $9 billion in 2025. The company has committed to substantial cloud infrastructure investments to support expanding enterprise usage. Regulatory developments include a $1.5 billion copyright settlement and U.S. government restrictions on foreign access to advanced models.

Additional features introduced include a Fast mode at 2.5 times speed for double the price, automatic fallback routing on the API, and mid-conversation tool changes that preserve prompt cache. Opus 5 is available immediately via the Claude API as claude-opus-5, with no data retention requirements for general access.

Anthropic’s strategy reflects a broader industry shift from showcasing peak AI capabilities to delivering reliable, cost-efficient models for everyday enterprise applications. The success of Opus 5 will depend on its performance in real-world workloads and enterprise acceptance of safety classifiers managing model responses.

Key topics in this update include anthropic, claude opus, and cheaper ai model.