Anthropic Launches Claude Opus 5 for Advanced AI Tasks

AI News

July 25, 2026

Anthropic Launches Claude Opus 5 for Advanced AI Tasks

Anthropic has launched Claude Opus 5, its latest high-performance artificial intelligence model for coding, research, financial analysis and long-running agent workflows.

The new model arrives less than two months after Opus 4.8 and is positioned below Anthropic’s more powerful Fable 5 system. However, Opus 5 is cheaper, has fewer access restrictions and outperforms Fable 5 on several benchmarks included in the launch announcement.

Stronger coding and reasoning

Anthropic says Opus 5 is more reliable when working on complex tasks that require repeated testing, verification and revision.

The model is designed to check its own work more carefully instead of stopping after producing an initial answer. This could make it more useful for software engineering, codebase migrations, financial modelling, document analysis and autonomous agent workflows.

Early customers reported improvements in long-horizon coding, numerical reasoning, legal analysis, data work and professional research. Some testers also found that Opus 5 completed difficult tasks with fewer tool calls and less time than Opus 4.8.

Large context window

Claude Opus 5 supports a one-million-token context window and can generate outputs of up to 128,000 tokens.

The large context capacity allows developers to provide extensive code repositories, documents or conversation histories in a single request. The model also uses extended thinking by default, with adjustable effort levels ranging from low to max.

These controls allow developers to balance speed and cost against deeper reasoning. Lower effort settings can be used for routine work, while higher levels are intended for difficult coding, research and agent tasks.

Pricing stays unchanged

Anthropic kept Claude Opus 5 API pricing at $5 per million input tokens and $25 per million output tokens, the same rates charged for Opus 4.8.

Maintaining the previous price makes the new model easier for existing customers to adopt without changing their basic cost assumptions. Opus 5 is available through the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry.

TechCrunch noted that Opus 5 may become the preferred Anthropic model for many professional workloads because it offers strong performance without some of the limitations attached to Fable 5.

Fewer restrictions than Fable

Opus 5 is not covered by the special 30-day data-retention policy applied to Anthropic’s Fable and Mythos models.

The model still includes safeguards, particularly for cybersecurity tasks. It can help identify vulnerabilities in source code, but restrictions may block activities such as scanning compiled software binaries for exploitable weaknesses.

This approach gives developers broader access while preserving limits around higher-risk offensive security work.

A practical flagship model

Opus 5 fills the gap between Anthropic’s lower-cost Sonnet series and its most powerful Fable system.

Its combination of advanced reasoning, large context, broad cloud availability and unchanged pricing could make it attractive to businesses that need stronger AI performance without moving to Anthropic’s most restricted model.

The launch also completes most of Anthropic’s transition to the Claude 5 generation. Fable 5, Mythos 5 and Sonnet 5 were released in June, leaving only the lightweight Haiku line without a version 5 update.

Productivity