On February 25, Amazon-backed AI startup Anthropic officially launched its latest AI model—Claude 3.7 Sonnet—claiming it to be the "smartest" AI the company has ever developed. This unique "hybrid" model combines both a "reasoning mode" (which pauses to contemplate complex problems) and a traditional mode (which generates answers in real-time), making it the only model of its kind available on the market.

Key Features and Capabilities

Jared Kaplan, co-founder and Chief Scientist of Anthropic, stated in a media interview that this model integrates multiple functions, aiming to provide a unified AI solution adaptable to various application scenarios. He likened it to how the human brain operates: some problems require deep thinking, while others need quick responses. By merging these two capabilities within a single model, Anthropic hopes to simplify the user experience, allowing users to focus less on making choices.

Claude 3.7 Sonnet excels particularly in the programming domain, capable of handling complex codebases and demonstrating impressive skills in using advanced tools. In multiple tests, this model was rated as the best programming model, especially in planning code changes and managing full-stack updates. Additionally, Claude’s accuracy in complex agent workflows has been recognized, showcasing its strong practicality.

User Experience and Feedback

User experience is crucial, and Claude 3.7 Sonnet allows users to finely control the thinking time through an API, setting a budget for thought (e.g., allowing up to N tokens). This enables users to balance speed and answer quality. One user noted that when they asked Claude to visualize accounting analysis data, it took only ten seconds, expressing amazement that “we might not even need to work anymore,” highlighting its efficiency and practicality.

However, not all feedback has been positive. Some users have raised concerns about Claude's knowledge base, noting that its information is limited to before October of last year. One user inquired about the current US president, and Claude incorrectly answered Kamala Harris, correcting itself only after being prompted that it should be Trump. Additionally, some users feel that the costs associated with using Claude are relatively high, especially when utilizing the reasoning function. The charges are $3 per million input tokens and $15 per million output tokens, which can lead to significantly higher overall costs, particularly when complex thinking is involved.

Overall, the launch of Claude 3.7 Sonnet signals an intensifying competition in the AI field, with Anthropic hoping to gain an advantage over competitors like OpenAI. As technology continues to advance, user expectations for AI models also rise. Kaplan indicated that more competitors are likely to follow this hybrid model approach, and the rapid development of the AI industry will drive further innovation.

How to Trade on uSMART

After logging into the uSMART HK APP, click on the search icon at the top right of the screen. Enter the stock code, such as "AMZN.US" to access detailed information, trading history, and trends. Click the “Trade” button at the bottom right, select the “Buy/Sell” function, and submit your order after filling in the transaction conditions.

（Source: uSMART HK）