Anthropic has announced its latest Claude 4 Model family. This seems like a leap for anyone building next-generation AI assistants or coding. The show’s stars are Claude Opus 4, The New Powhouse, and Claude Sonnet 4, which are designed to be smart all-rounders.
Humanity is not shy about its ambitions, saying that these models aim to “follow customers’ AI strategies across the board.” While Sonnet 4 has placed the Opus 4 as a tool “pushing the boundaries of coding, research, writing and scientific discovery” while being billed as a “instant upgrade from Sonnet 3.7” as a “returning frontier performance to everyday use cases.”
Claude Opus 4: New Coding Champion
When humanity calls the Claude Opus 4 “the most powerful model ever and the best coding model in the world”, you sit down and notice. And they have the numbers to back it up, Opus 4 topped the charts in key industry tests, reaching 72.5% on the SWE bench and 43.2% on the terminal bench.
But it’s not just a sprint. The Opus 4 is built for long distances designed for “sustainable performance of long-term tasks that require focused effort and thousands of steps.” Imagine an AI that can “work continuously for several hours.”
This should be a major step up from the previous sonnet model, allowing us to expand what AI agents can achieve and tackle problems that require actual persistence.
Claude Sonnet 4: For daily AI and agent work
The Opus 4 is a heavyweight champion, but the Claude Sonnet 4 has become a versatile mainstay, significantly boosting the vast range of applications. Early feedback from those who had sneak peaks shines.
For example, Github said “Claude Sonnet 4 will soar in the agent scenario,” and was very impressed with “planning to present it as the base model for Github Copilot’s new coding agent.” That’s a huge amount of support.
Technology commentator Manus was also impressed, highlighting “complex instructions, clear reasoning, and improved aesthetic output.”
The positive atmosphere reports that “Sonnet 4 is excellent at autonomous multi-feature app development, and that it has a significant improvement in problem solving and codebase navigation, reducing navigation errors from 20% to near zero.” It is a game changer in development workflows.
SourceGraph is equally optimistic, viewing the model as “a substantial leap in software development – running longer, a deeper understanding of the issues, and providing more elegant code quality.”
Augment Code shows “high success rates, more surgical code editing, and more careful work through complex tasks,” leading Sonnet 4 to “top selection of primary models.”
Hybrid Mode and Developer’s Joy
One really clever bit about Claude 4’s family is its hybrid nature. Both the Opus 4 and Sonnet 4 can operate in two gears. One is the close-close replies we often need, and the other allows for “extended thinking for deeper reasoning.”
This deeper mode of thinking is part of the Pro, Max, Teams, and Enterprise Claude Plan. But the good news for everyone is that Sonnet 4 will be available to free users with this expanded thinking. This is a great move to make Top Tier AI more accessible.
Humanity is also deploying delicious new tools for developers in its APIs, and clearly aims to enhance the creation of more sophisticated AI agents.
- Code Execution Tool: This allows the model to actually execute code, opening up all kinds of possibilities for interactive and problem-solving applications.
- MCP Connector: Introduced by humanity, MCP standardizes the exchange of contexts between AI assistants and software environments.
- File API: This makes it much easier for AI to directly link files. This is a big deal for many real tasks.
- Prompt caching: Developers can cache prompts for up to an hour. This may sound small, but it can make a huge difference in speed and efficiency, especially in frequently used queries.
Leading the pack in real performance
Humanity is keen to highlight “the benchmark of performance for real software engineering tasks, where the Claude 4 models lead the SWE-Bench Verified.” Beyond coding, these models emphasize that they “provides powerful performance across coding, inference, multimodal features, and agent tasks.”
Despite the leaps in capabilities, humanity holds a line of pricing. Claude Opus 4 returns $15 per million input tokens and $75 per million output tokens. A more accessible option, the Claude Sonnet 4 costs $3 per input token and $15 per million output token. This consistency is welcomed by existing users.
Both Claude Opus 4 and Sonnet 4 are ready to go through human APIs, and are also appearing in Vertex AI on Amazon Bedrock and Google Cloud. This wide availability means that businesses and developers around the world can begin experimenting and integrating these new tools fairly easily.
Humanity is clearly doubling its AI more capable, especially in the complex areas of coding and autonomous drug behavior. These new models and developer tools have provided a serious boost to the potential for innovation.
(Image credit: Humanity)
reference: Jony Ive’s Ambitious Openai Device Details Leak
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London. The comprehensive event will be held in collaboration with other major events, including the Intelligent Automation Conference, Blockx, Digital Transformation Week, and Cyber Security & Cloud Expo.
Check out other upcoming Enterprise Technology events and webinars with TechForge here.