Content
summary Summary

Anthropic has taken a different approach from OpenAI with its latest release, introducing Claude 3.7 Sonnet as its first reasoning-enabled model that combines direct responses with extended thinking capabilities.

Ad

Claude 3.7 Sonnet builds on an upgraded version of Claude 3.5 Sonnet but adds a significant new feature: an "Extended Thinking" mode that allows the model to reason through its responses. According to Anthropic, reasoning should be an integrated capability of advanced models. Like OpenAI's reasoning models, Claude 3.7 Sonnet generates reasoning tokens. This release puts Anthropic ahead of OpenAI, which recently announced similar unified architecture plans for GPT-5.

Alongside the model update, Anthropic has introduced Claude Code, a command-line tool for agent-based programming. This tool allows developers to delegate complex programming tasks directly through their terminal. Claude Code is currently available as a limited research preview.

Users can choose when the model should provide immediate responses and when it should engage in extended thinking. In the extended thinking mode, the model reflects more deeply, improving its performance across mathematics, physics, programming, and many other tasks.

Ad
Ad

Through the API, users can control the "budget" in tokens, effectively managing the time and associated costs of the reasoning process. Anthropic says it focused on developing practical capabilities that reflect how businesses actually use LLMs.

Enhanced programming and debugging capabilities

According to Anthropic's initial testing, Claude leads the field in practical programming tasks, significantly outperforming other models in planning code changes and complex updates.

Vergleichstabelle: Leistungsmetriken von 7 KI-Modellen (Claude, OpenAI, DeepSeek, Grok) in 8 Kategorien wie Reasoning, Coding und Math.
In mathematical problem-solving and multilingual Q&A in particular, the model achieved top scores of over 90 percent, on a par with the o-models from OpenAI or DeepSeek-R1. | Image: Anthropic

The company reports that Claude can build sophisticated web apps and dashboards from scratch where other models struggle. In evaluations, Claude consistently produced production-ready code with superior design and dramatically fewer errors, according to the official blog post.

Claude Code: A new approach to agent-based programming

Claude Code, available as a limited research preview, represents Anthropic's first agent-based programming tool. It can search and read code, edit files, write and run tests, commit and push code to GitHub, and use command-line tools.

Initial tests show Claude completing tasks that typically require over 45 minutes of manual work in a single pass, significantly reducing development time and effort.

Recommendation

Anthropic plans to continuously improve the tool by enhancing the reliability of tool calls, supporting long-running commands, and improving app rendering. The company aims to better understand how developers use Claude for programming to guide future model improvements.

Improved safety and response handling

According to the company, Claude 3.7 Sonnet makes more nuanced distinctions between harmful and harmless requests, reducing unnecessary rejections by 45 percent compared to its predecessor. Anthropic has conducted extensive testing and worked with external experts to ensure the model meets safety and reliability standards.

The company has prioritized safe and responsible AI use from the beginning, though this sometimes leads to surprising and questionable limitations due to Claude's cautious approach to prompts.

Allegedly still 2 years until AI breakthroughs

Anthropic sees Claude 3.7 Sonnet and Claude Code as crucial steps toward AI systems that truly augment human capabilities. The company predicts that by 2027, Claude will find "breakthrough solutions to challenging problems." Currently, it claims to save "hours of work" by performing at expert-human levels.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.

The company has also highlighted Claude's potential for agent-based tasks, recently demonstrating this with Claude Computer Use, which enables AI-supported control of software and operating systems.

Zeitleiste 2024-2027: Evolutionsstufen von Claude - von
This is how Anthropic predicts the medium-term future for Claude. | Image: Anthropic

Anthropic appears to be focusing primarily on the mid-tier Sonnet series. While Sonnet and the smaller Haiku were recently updated to version 3.5, the largest model, Opus, remains at version 3.

Anthropic's enterprise growth strategy

The release of Claude 3.7 Sonnet and Claude Code aligns with Anthropic's ambitious growth strategy. The company aims to reach revenue of up to $34.5 billion by 2027, a significant increase from the projected $3.7 billion in 2025.

Anthropic particularly wants to surpass OpenAI in the lucrative API business, targeting triple OpenAI's revenue by 2027. The company expects to become profitable this year.

Claude 3.7 Sonnet is now available across all Claude plans, from free to enterprise versions, as well as through the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI. The extended thinking mode is available on all platforms except the free Claude version, with pricing remaining unchanged.

Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • Anthropic launches Claude 3.7 Sonnet, a hybrid AI model capable of both rapid response and step-by-step detailed reasoning. Users can choose when the model should respond normally and when it should think longer, improving performance on complex tasks.
  • Claude 3.7 Sonnet demonstrates leadership in real-world programming tasks in tests. It consistently produces production-ready code with superior design and reduced errors. The Claude Code command line tool allows developers to delegate large programming tasks to Claude directly from the terminal.
  • The release is part of Anthropic's ambitious growth strategy to reach up to $34.5 billion in revenue by 2027. Especially in the lucrative API business, Anthropic aims to overtake competitor OpenAI and is relying on Claude's advanced reasoning and programming capabilities.
Sources
Jonathan writes for THE DECODER about how AI tools can make our work and creative lives better.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.