The latest versions of Anthropic’s Claude generative AI models made their debut Thursday, including a heavier-duty model built specifically for coding and complex tasks.

Anthropic launched the new Claude 4 Opus and Claude 4 Sonnet models during its Code with Claude developer conference and executives said the new tools mark a significant step forward in terms of reasoning and deep thinking skills.

The company launched the prior model, Claude 3.7 Sonnet, in February. Since then, competing AI developers have also upped their game. OpenAI released GPT-4.1 in April, with an emphasis on an expanded context window, along with the new o3 reasoning model family. Google followed in early May with an updated version of Gemini 2.5 Pro that it said is better at coding.

Claude 4 Opus is a larger, more resource-intensive model built to handle particularly difficult challenges. Anthropic CEO Dario Amodei said test users have seen it quickly handle tasks that might have taken a person several hours to complete. 

«In many ways, as we’re often finding with large models, the benchmarks don’t fully do justice to it,» he said during the keynote event.

Claude 4 Sonnet is a leaner model, with improvements built on Anthropic’s Claude 3.7 Sonnet model. The 3.7 model often had problems with overeagerness and sometimes did more than the person asked it to do, Amodei said. While it’s a less resource-intensive model, it still performs well, he said. 

«It actually does just as well as Opus on some of the coding benchmarks, but I think it’s leaner and more narrowly focused,» Amodei said.

Anthropic said the models have a new capability, still being beta tested, in which they can use tools like web searches while engaged in extended reasoning. The models can alternate between reasoning and using tools to get better responses to complex queries.

The models offer near-instant response modes and extended thinking modes. 

All of the paid plans offer Opus and Sonnet models, while the free plan just has the Sonnet model.

The new models show Anthropic’s focus on building strong coding models, said Arun Chandrasekaran, a distinguished vice president, analyst at Gartner. «Anthropic’s Claude models have established strong leadership in the software engineering domain and the latest Claude 4 release extends that leadership.»

Anthropic triggers safety protocols with new Claude models

In launching the Claude Opus 4 model, Anthropic said it was taking increased safety precautions to reduce the risk of Claude being misused. In a blog post, the company said it hasn’t determined whether the model actually requires the protections of its ASL-3 standard but it is doing so as a precaution. 

The safety precautions are specifically designed to prevent Claude from helping with developing chemical, biological, radiological or nuclear weapons. Anthropic said it limited attacks known as universal jailbreaks that let attackers get around existing protocols.  «We will continue to evaluate Claude Opus 4’s CBRN capabilities,» Anthropic’s blog post said. «If we conclude that Claude Opus 4 has not surpassed the relevant Capability Threshold, then we may remove or adjust the ASL-3 protections.»

Chandrasekaran said the implementation of safety standards is worth noting. «This includes enhanced cybersecurity measures and prompt classifiers to mitigate risks associated with powerful AI systems,» he said. The new models show the company’s focus on balancing new technology with safety, he said.