What Is Claude Keep Thinking | Extended Thinking Feature Explained

What Is the "Claude Keep Thinking" Campaign?

In 2025, Anthropic launched its first paid brand campaign, "Keep thinking," across the United States. Centered on a 90-second film produced by independent agency Mother London, it ran on major media outlets including Netflix, Hulu, the New York Times, and the Wall Street Journal. Media agency Initiative handled the buy, and investment was reported to be in the multi-million-dollar range (source: creative.salon).

The campaign's message is clear: "AI does not replace human thinking — it deepens human thinking as a partner." Under this concept, the campaign outlines a positioning strategy that differentiates Claude as a "problem-solving AI" from competitors.

The impetus for Anthropic investing in paid advertising at this moment is the acceleration of business growth. The number of business customers, which was under 1,000 two years ago, has surged to more than 300,000. Establishing a strong brand image became urgent to build awareness and trust among enterprise buyers.

Why "Keep Thinking" Is Anthropic's Message

The message "Keep thinking" is more than a tagline. It reflects the philosophy Anthropic places at the center of how it designs Claude.

While other AI tools compete to "produce answers instantly," Anthropic argues for "valuing the process of thinking." This contrast holds particular appeal for business users. Complex decision-making, strategic planning, and creative problem-solving derive value from deep thinking processes rather than instant answers.

This stance is also clearly articulated in Anthropic's official statement, "Claude is a space to think" (source: anthropic.com). Anthropic declared that Claude will carry no advertising whatsoever, clearly staking out the position of "a space for thinking where users can comfortably share personal and confidential information." Monetization is limited solely to enterprise contracts and paid subscriptions, ensuring transparency in recommendations.

What Is Extended Thinking — Claude's "Visible Reasoning" Feature

The "Keep thinking" philosophy is embodied in a concrete Claude API feature: Extended Thinking.

Extended Thinking is a feature that executes and displays a step-by-step reasoning process before Claude produces its final answer (source: Extended Thinking official documentation). Ordinary AI responses resemble a black box — input goes in, an answer comes out — but with Extended Thinking enabled, you can follow "what Claude is thinking and how" in real time.

This feature offers two major benefits.

Improved answer reliability: For complex math, logic, and coding problems, making reasoning steps explicit reduces errors and produces more accurate answers.

Process transparency: Users can trace Claude's reasoning, making it easy to evaluate critically and make corrections. This is especially valuable in business and research contexts.

The situations where Extended Thinking is particularly effective can be summarized as follows.

How to Implement the Extended Thinking API

Extended Thinking is available in the Claude API and is controlled via the thinking parameter. Basic usage is as follows.

import anthropic

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-opus-4-7-20251101",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 10000  # 内部推論に使うトークン上限
    },
    messages=[{
        "role": "user",
        "content": "この経営課題を解決する最良の戦略は何ですか？"
    }]
)

budget_tokens sets the upper limit on tokens used for internal reasoning. A larger value produces answers that reflect deeper thinking, but also increases token consumption. Balancing this against the use case is the fundamental approach.

The supported models and their output token limits are as follows.

Note that Extended Thinking is covered under Zero Data Retention (ZDR), so it can be used confidently in environments that must meet corporate security requirements.

Adaptive Thinking — Automatic Optimization in Claude Opus 4.7

Extended Thinking has evolved further with the arrival of "Adaptive Thinking."

In the conventional enabled mode, users had to set budget_tokens manually. With Adaptive Thinking, Claude judges the complexity of the task itself and automatically adjusts the amount of thinking.

response = client.messages.create(
    model="claude-opus-4-7-20251101",
    max_tokens=16000,
    thinking={
        "type": "adaptive"  # モデルが思考量を自動判断
    },
    messages=[...]
)

Simple questions receive brief reasoning; complex problems get deep inference. The biggest benefit is that users are freed from tuning token costs and can focus entirely on the task at hand.

There is also a feature called "Interleaved Thinking." This mode alternates between tool calls and reasoning, enabling a continuous process of "fetching information from the web, reflecting on its content, then deciding the next action." It is activated via the beta header interleaved-thinking-2025-05-14.

Claude as a "Space to Think" — The Strategic Meaning of the No-Ads Declaration

In the statement "Claude is a space to think," Anthropic made clear that Claude will not be used as an advertising medium.

This declaration is also strategically significant as a differentiator from competitors. AI services that run on an advertising model have an incentive to optimize user behavioral data and interests for advertisers — something fundamentally at odds with the goal of "supporting users' thinking."

Claude's monetization is limited to the following models:

• Max plan, Team plan, Enterprise plan: Paid subscriptions
• Enterprise integrations: Third-party integrations with Figma, Asana, Canva, and others
• Educational support: Programs for educators in 60+ countries, national AI pilots with governments around the world

By maintaining a no-advertising policy, Claude can offer "the answers users truly need" free of commercial bias. Building an environment where users can comfortably share personal and confidential information is the prerequisite for "Keep Thinking."

Practical Ways to Make the Most of Keep Thinking

To get the most out of the "Keep Thinking" philosophy and its features, Extended Thinking is most effective in the following situations.

Assisting with complex decision-making

For multi-variable problems such as business strategy or investment decisions, visualizing Claude's thinking process makes it easier to spot overlooked perspectives and verify underlying assumptions. Because you can see "how it arrived at that conclusion" — a process typically omitted in ordinary responses — you can critically evaluate rather than accept the answer at face value.

Code review and debugging

Enabling Extended Thinking when identifying problematic code lets you see "which logic Claude traced to find the issue," making it easier to understand the root cause yourself. Rather than simply receiving "fixed code," you can use the debugging thought process as a learning resource.

Academic and research use

Combining Learning mode with Extended Thinking enables a Socratic style of dialogue. Instead of handing you the answer directly, Claude deepens your understanding by showing its reasoning process. It is especially useful as a supplement for peer review and hypothesis testing.

Prompt engineering optimization

Observing Claude's internal reasoning process lets you experimentally optimize "which prompts are effective." Comparing the differences in reasoning paths when you change inputs leads to better prompt design.

Conclusion — Why It Matters That Claude "Keeps Thinking"

"Claude Keep Thinking" is both the message of a campaign Anthropic is running and a concrete feature of the Claude API.

The brand campaign "Keep thinking" positions AI not as a "replacement for thinking" but as a "partner in thinking," conveying Claude's value to more than 300,000 business customers. On the technical side, Extended Thinking and Adaptive Thinking make it possible to visualize and control Claude's internal reasoning process. This provides particularly great value to professional users tackling complex problems.

And the no-ads declaration is an expression of Claude's commitment to "prioritizing users' thinking above all." Together, these elements form the position Anthropic is aiming for: "Claude = a space for thinking."

When facing a complex challenge, rather than settling for an instant answer, using Claude to "keep thinking together" is what draws out the true value of this AI.

Sources: Anthropic "Keep thinking" campaign details / Extended Thinking official documentation / Claude is a space to think