Anthropic has made its artificial intelligence model Claude Opus 4.8 available to users. The company increased its development speed by releasing this new version just six weeks after the previous model update.

The new model draws attention with its sharper judgment ability and longer independent working capacity compared to previous versions. Anthropic notes that Opus 4.8 is more honest about its progress and less likely to make unsupported assertions.

Improved performance and new capabilities

The performance increase that comes with Claude Opus 4.8 provides measurable results in various areas. While the agentic coding score increases from 64.3 percent to 69.2 percent, multidisciplinary reasoning ability increases from 54.7 percent to 57.9 percent.

While the knowledge process score reaches 1890 from 1753, financial analysis skills also increase to 53.9 percent. The company states that the fast mode of Opus 4.8 works approximately 2.5 times faster and the cost of this mode is three times lower.

With the updates made to Claude Code, users can now choose extra and maximum performance settings. These settings allow the model to spend more tokens to produce better results in more complex missions.

Anthropic recommends using the extra setting for powerful missions and long-running asynchronous workflows. In addition, the speed limits on the Claude Code have been increased to support these high performance levels.

Dynamic workflows and new API features

With the new model, the dynamic workflows feature, which is in the research-based preview phase, is available. This feature allows Claude to address much larger scale problems within Claude Code.

Users can now determine how much effort the model will spend on a response through a new control panel available on claude.ai and Cowork platforms. The Messages API now accepts system input within the message thread.

In this way, developers can update instructions during the task without breaking the prompt cache or redirecting the user. Anthropic also plans to offer its cybersecurity-focused model called Mythos to all its customers in the coming weeks.

The Mythos model, which was announced in April, could only be used by stakeholders on certain software platforms until now. Claude Opus version 4.8 is available worldwide as of today.

What do you think will change the world of artificial intelligence if Anthropic releases a new model in such a short time?