Claude 4 Sonnet, Opus AI models released with enhanced coding capabilities

Anthropic claims Claude Opus 4 can compete with GPT-4.1 and Gemini 2.5, while Sonnet 4 outperforms its predecessor in reasoning and code generation

Anthropic has announced the launch of its new Claude 4 Sonnet and Claude 4 Opus AI models, now available through Claude’s website and in Application Programming Interface (API). According to the company’s official blog post, both models bring significant improvements in coding abilities, with Claude 4 Opus in particular targeting state-of-the-art performance across a variety of AI benchmarks.

 

The company stated that Claude Sonnet 4 outperforms its predecessor, Sonnet 3.7, while Claude Opus 4 is said to match or exceed the capabilities of competing large language models like OpenAI’s o3, GPT-4.1, and Google’s Gemini 2.5 Pro. These results are based on benchmarks that test areas such as multilingual proficiency, agentic tool use, autonomous coding via terminal interfaces, and graduate-level reasoning.

 

 

The announcement comes just weeks after both OpenAI and Google released updated AI models with enhanced coding capabilities. OpenAI debuted its o3 and GPT-4.1, while Google launched Gemini 2.5 Pro, which drew attention after reportedly completing a full playthrough of Pokémon Blue.

 

In comparison, Anthropic shared that while Claude Opus 4 has not yet completed the game, it was able to autonomously play for 24 hours straight—a significant improvement over its predecessor, which managed only 45 minutes of autonomous gameplay.

 

New capabilities and Claude Code

Anthropic has also announced some new capabilities. Here are the details:

• Extended reasoning with tools (beta): Claude models can now use tools like web search while engaging in deeper reasoning. This lets them switch between thinking and tool use to generate better results.
• Improved model abilities: Claude 4 Sonnet and Claude 4 Opus can now run tools simultaneously, follow instructions with greater accuracy, and, when granted access to local files, show enhanced memory—picking up and storing key information to ensure better context retention and learning over time.
• Claude Code is now widely available: Claude Code is now open for broader use. It supports background processes through GitHub Actions and integrates directly with development tools like VS Code and JetBrains, offering in-line edits for smoother collaborative coding.

• New API features: Anthropic has added four new tools to its API: a code execution feature, an MCP connector, a Files API, and prompt caching for up to one hour—allowing developers to create more capable AI-driven applications.

Claude 4 Sonnet and Opus: Availability

Anthropic states that both models are accessible via the Anthropic API, as well as through partners like Amazon Bedrock and Google Cloud’s Vertex AI.

 

Pricing for Opus 4 is set at $15 per million input tokens and $75 per million output tokens, while Sonnet 4 is priced at $3 for input and $15 for output per million tokens.