Anthropic 'Claude 3.7 Sonnet' And 'Claude Code' Mark The Next Step Of Human-LLM Collaboration

What is one but not the other. Sometimes, requirements differ from one people to another.

Since OpenAI announced ChatGPT and sparked an arms race between tech companies of all sizes, Anthropic emerged as one of the most powerful competitors. Its Claude 3.5 Sonnet.

This Large Language Model is able to competes with OpenAI GPT-4o, but with a better sense of humor.

And just as many of its rivals have introduced their own reasoning models, Anthropic came up with 'Claude 3.7 Sonnet,' which can engage in "extended thinking."

Similar to reasoning models from rivals like OpenAI o3 and DeepSeek-R1, Claude 3.7 Sonnet can "reason" through challenging problems by questioning itself, apply more computing, and take more time to verify its thoughts.

But Anthropic is tweaking that ability a bit

Instead of making two different models to serve two different purposes, the switch allows Claude 3.7 Sonnet AI model to turn on/off its 'Extended Thinking Mode.'

Anthropic describes Claude 3.7 Sonnet as "both an ordinary LLM and a reasoning model in one."

At the front, it boasts a significant upgrade over its predecessor, and that it can produce near-instant responses. But when the mode is switched on, Claude 3.7 Sonnet can instantly benefit from an additional boost in math, physics, instruction-following, coding, and many other tasks.

With this switch, it's up to the users to decide whether or not the model should generate a quick response or engage in a deeper reasoning process.

For transparency, step-by-step thinking is made visible to the user. This allows users to see the whole process of the AI's thoughts, allowing users to verify answers and detect inconsistencies.

But the whole idea of this switch, is to give developers the ability to set a "thinking budget," controlling how much time and computational resources Claude is allowed to allocate to problem-solving. In API applications, they can limit the number of tokens used for extended reasoning up to a maximum of 128K tokens.

By giving developers the ability to control over how long the model can think for, the approach allows for a trade-off between response speed, cost, and output quality.

Unlike switching to an entirely different model, this feature enhances the same model's ability to perform deeper reasoning.

Available across all Claude plans, including Free, Pro, Team, and Enterprise, as well as through Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI, Claude 3.7 Sonnet uses a "different philosophy than other reasoning models out there."

In the announcement, Anthropic said that:

This is why Claude 3.7 Sonnet is "the first hybrid reasoning model on the market," said Anthropic.

Anthropic explained that Claude 3.7 Sonnet embodies the philosophy by being both an ordinary LLM and a reasoning model, all at the same time.

If users want it to answer normally, it's up to them. Or if they want it to think longer, they can do that easily.

In the standard mode, Claude 3.7 Sonnet is essentially an upgraded version of Claude 3.5 Sonnet. And in extended thinking mode, its reasoning model kicks in, allowing it to self-reflects before answering.

The company benchmarked Claude 3.7 Sonnet by playing Pokémon Red, by equipping the model with basic memory, screen pixel input, and function calls to press buttons and navigate the game.

This setup allowed it to play continuously beyond standard context limits, sustaining gameplay through tens of thousands of interactions.

Unlike the previous version of Claude which failed to leave the house in Pallet Town where the story begins, Claude 3.7 Sonnet's reasoning ability made it able to battle three Pokémon gym leaders and won their badges.

In addition to Claude 3.7 Sonnet, the company is also launching 'Claude Code,' which is its first "agentic coding tool."

Initially available in a limited research preview, Claude Code is an active collaborator that can search and read code, edit files, write and run tests, commit and push code to GitHub, and use command line tools.

What this means, Claude Code enables developers to delegate substantial engineering tasks to Claude directly from their terminal.

Anthropic claims that its team has been using Claude Code, and has been making it "indispensable for our team."

From test-driven development, debugging complex issues, and large-scale refactoring, early tests reveal that Claude Code is able to complete tasks in a single pass that would normally take 45+ minutes of manual human work.

This, according to Anthropic, should reduce development time and overhead.