Anthropic Unveils Claude 4 Models, Intensifying AI Competition

Ticker BuzzThursday, May 22, 2025 6:03 pm ET
2min read

Anthropic, a prominent competitor to OpenAI, has unveiled its latest series of models, Claude 4, which includes two new models: Claude Opus 4 and Claude Sonnet 4. The company has dubbed Opus 4 as the "world's best programming model," marking a new era in model capabilities and potentially reshaping the software development industry.

In addition to the new models, Anthropic has announced the official launch of its programming tool, Claude Code. This tool offers new functionalities for developers connected to its models, including the ability to integrate reasoning with web searches and other tools. This launch intensifies the competition with OpenAI and Google in the top-tier model domain, providing investors with an opportunity to reassess the competitive landscape in the AI sector.

Opus 4 is described as Anthropic's most powerful model to date, excelling in coding, research, writing, and scientific discovery. Sonnet 4, an upgrade from Sonnet 3.7, offers superior coding and reasoning capabilities while responding more accurately to user commands. Both models feature a hybrid design that allows for nearly instantaneous responses as well as deeper, more expansive thinking.

Opus 4 and Sonnet 4 are included in Anthropic's Pro, Max, Team, and Enterprise Claude packages, with Sonnet 4 also available to free users. These models can be accessed through Anthropic API, Amazon Bedrock, and Google Cloud's Vertex AI. The pricing for Opus 4 is 15 dollars per million tokens for input and 75 dollars for output, while Sonnet 4 costs 3 dollars per million tokens for input and 15 dollars for output.

One of the standout features of Opus 4 is its remarkable ability to maintain performance over extended periods. Anthropic reports that the model can sustain stable performance for several hours, even in tasks requiring thousands of steps. This capability was validated by the Japanese e-commerce giant Rakuten, where Opus 4 ran independently for 7 hours on a demanding open-source reconstruction task, maintaining stable performance throughout.

This sustained focus and work capability make Opus 4 particularly suitable for complex coding and problem-solving scenarios. Code editing platforms like Cursor and Replit have praised Opus 4 for its significant improvements in understanding complex codebases and making intricate changes across multiple files.

Both new models introduce an "extended thinking and tool usage" feature, allowing Claude to use tools like web searches during the thinking process. This capability enhances the models' ability to follow instructions more precisely and improves their memory when granted local file access by developers. Opus 4, in particular, has shown significant advancements in memory capabilities, creating and maintaining "memory files" to store key information, which enhances long-term task awareness and performance.

With the launch of Claude Code, Anthropic aims to revolutionize AI-driven development. Supported by Opus 4 and Sonnet 4, Claude Code integrates seamlessly with GitHub Actions, VS Code, and JetBrains, enabling real-time editing and pair programming. This tool allows developers to analyze data more effectively, as Claude can write and debug code without the user's knowledge, leveraging its coding capabilities to process large datasets.

Scott White, the product manager at Anthropic, highlighted the growing and exciting market for coding, noting that Claude Code is designed for both professional software engineers and non-programmers. For instance, a product manager can use Claude to create a prototype of an idea, rather than just documenting the concept.

The release of Claude 4 comes at a time of intense activity in the AI sector, with other major players like Microsoft and Google also making significant announcements. This heightened competition underscores the need for investors to reassess the industry landscape, particularly the positioning of Anthropic relative to its rivals. The strong performance of Claude 4 in coding, reasoning, and proxy tasks could provide Anthropic with opportunities to gain more market share and enterprise clients.