Anthropic has just launched Claude 4, introducing two new models: Sonnet 4 and Opus 4. According to the company, Opus 4 is a standout, described as “the world’s best programming model.” This claim is backed by its impressive performance on the SWE-Bench benchmark, where it scored 72.5%, slightly edging out OpenAI’s Codex (72%) and surpassing Gemini 2.5 Pro (63.2%).
One of Opus 4’s remarkable feats is its ability to refactor code continuously for seven hours without losing coherence, showcasing its potential for complex programming tasks. While Sonnet 4 is available to all users, access to the more advanced Opus 4 requires a subscription.
These details come from a recent report by Ars Technica, highlighting Anthropic’s continued push to advance AI capabilities for developers and businesses alike. As AI models evolve, Claude 4’s performance sets a new benchmark for what’s possible in automated programming.
Stay tuned for more updates on AI innovations and their impact on the tech world!
Source: Ars Technica
