Anthropic is preparing to launch two upgraded versions of its Claude AI software, including a much-anticipated update to its premium Opus model, as it competes in an increasingly crowded AI landscape.
On Thursday, the company will introduce Sonnet 4 and Opus 4—its most advanced AI system to date. These new models more accurately follow instructions and independently handle tasks such as coding and answering complex queries.
Founded in 2021 by former OpenAI employees, Anthropic has aimed to set itself apart with more advanced AI models and a strong focus on ethical development. Despite this mission, the company—like many of its competitors—faced challenges last year in delivering AI systems that offered significant performance gains relative to their high development costs.
In March 2024, Anthropic released three models, including Opus 3. Anthropic rolled out updates to the other two Claude models in the following months but left the Opus model unchanged. According to news reports, internal testing revealed that the new version, Opus 3.5, did outperform its predecessor but not to the extent expected based on its increased model size.
Ultimately, Anthropic decided against releasing the 3.5 version of its Opus model, even though it had earlier indicated a 2024 launch. Mike Krieger, Anthropic’s Chief Product Officer, said the company believed the updated Opus model didn’t offer significant enough improvements over the Sonnet model to justify a release or create meaningful differentiation.
“We’ve had large models trained internally. We’ve used them even internally for other applications,” Krieger said. “But we wanted the next one that we released publicly to feel like it is that leap.”
For Opus 4, Anthropic’s research team concentrated on enhancing the model’s ability to operate autonomously for extended periods while maintaining awareness of its ongoing tasks, according to Chief Product Officer Mike Krieger. In early testing, Japanese e-commerce giant Rakuten Group Inc. successfully used Opus 4 for seven continuous hours to refine open-source code—demonstrating its improved performance and long-term task handling capabilities.
“In the past, the model would be able to come up with a plan that was a several hour plan,” Krieger said, but eventually it would “get stuck somewhere” or lose its coherence.
Both Opus 4 and Sonnet 4 come with a relatively new feature that allows users to choose between receiving a quick, straightforward answer or a more in-depth, time-consuming response that simulates human-like reasoning.
With the launch of Opus 4 and Sonnet 4, the AI startup is signaling a significant leap forward in AI performance, autonomy, and user customization.
