Anthropic brings with Claude 3.7 Sonnet a new model that features a comprehensive thought process.
Anthropic has launched a “groundbreaking AI model” with Claude 3.7 Sonnet. The hybrid AI reasoning model possesses extensive thinking capabilities and thus can provide both quick answers and slightly longer thinking on complex problems.
Practical reasoning
“The model itself should recognize when a problem requires more intensive thinking to adapt accordingly, rather than a user choosing different modes of reasoning,” Dianne Penn, head of research at Anthropic, told VentureBeat.
Claude 3.7 Sonnet shows improvements in coding and front-end Web development. Companies report that Claude is effective in building sophisticated Web applications. Integration with GitHub enhances Claude’s usability as a software development tool.
The version of Claude 3.7 Sonnet without extended thinking performs slightly better compared to Claude 3.5 Sonnet, while the extended thinking mode is noticeably better in understanding text prompts and mathematics. Although DeepSeek R1 scores highest in math.
The new model is said to be tremendously secure, according to Anthropic. The start-up claims that Claude 3.7 Sonnet makes a better trade-off between harmful and harmless requests, which should ensure 45 percent fewer unnecessary rejections than its predecessor.
Anthropic also brings Claude Code on the market, a new tool for agent-based coding. This tool has limited availability and can read code, edit files, push code to GitHub and use command line tools to understand detailed commands.
The new Sonnet model is available on all Claude subscriptions (Free, Pro, Team and Enterprise) and through platforms such as Amazon Bedrock and Google Cloud’s Vertex AI.
read also