After weeks of speculation, AI startup humanity introduced the Claude 3.7 Sonnet, an AI model that combines quick responses with controlled, extended inference. In addition, the company has deployed Claude Code, a command-line coding tool designed to simplify development workflows. The launch is announced to show that humanity will secure a fresh $3.5 billion in funding and drive more advanced AI development.
“Introducing Claude 3.7 Sonnet: the most intelligent model ever. In an X post, Anthropic said, this is a hybrid inference model that produces a response that is closer to instant or extended, step-by-step thinking.” It states.
Introducing Claude 3.7 Sonnet: The most intelligent model ever. This is a hybrid inference model, which generates responses that are closer to instant and has a step-by-step approach.
One model, two ways to think.
We have also released the agent coding tool Claude Code. pic.twitter.com/jt7qqmfwuc
– Humanity (@anthropicai) February 24, 2025
A new approach to AI reasoning
Claude 3.7 Sonnet offers the option for users to switch between standard and extended thinking modes, allowing for more detailed inference when needed. API users can fine-tune the duration when the model processes information by supporting up to 128k tokens, allowing them to easily balance speed, cost and output quality.
Claude 3.7 Sonnet takes a major leap from its predecessor and introduces an extended thinking mode that hones your performance in mathematics, physics, coding and complex problem solving.
API users now have fine control over how long the model processes information, allowing them to balance speed, accuracy and cost based on their task requirements.
This model outperforms competitors such as the O1, O3-Mini and DeepSeek R1 in real-world coding benchmarks and using AI-driven tools. Available through the Claude app, Anthropic API, Amazon Bedrock, and Google’s Vertex AI. All of these are at the same price as their predecessors.
In a knowledge base updated as of October 2024, Claude 3.7 Sonnet is ideal for coding, finance and legal analysis, providing businesses with more capable AI to tackle complex tasks .
Claude Code: Hands-Free AI Assistant for Developers
Humanity also introduced Claude Code, a command-line assistant currently in a limited research preview. This tool can be done without all the need for file editing, code reading, test execution, and GitHub commit management, without the need for manual intervention. The goal is to reduce boring development tasks and allow engineers to focus on high-level problem solving.
Agent Coding
Claude code represents a shift towards agent coding. Here, AI doesn’t just propose code, it also starts to execute tasks. The tools are:
Read and understand existing code to make the exact changes to edit the file without the need for manual intervention execution tests.
By automating these iterative processes, Claude code reduces manual workloads for developers and allows them to focus on solving more complex problems.
Currently, limited research previews provide a glimpse into the future where humanity can push AI-driven development deeper and enable AI to act as independent coding assistants.
A fierce competitive landscape
AI-driven reasoning is becoming the next major battlefield, with humanity moving with Claude 3.7 Sonnet. The company has been relatively quiet in recent months, but this release brings it back to the spotlight. The race gets hot as Deepseek pushes open source models and Openai tackles its own hybrid AI.
Claude 3.7 Sonnet shows the shift in humanity. It is a more strategic, high-performance AI designed to allow users to have stronger control over inference and problem solving. The next generation of AI models are here, and the competition is just beginning.
This is a YouTube video from humanity covering the release of Claude 3.7 Sonnet and Claude Code.
