Spend any time in AI circles and you'll hear "Claude Labs," "Anthropic," "Claude," and "Claude Code" used almost interchangeably. None of these terms are wrong, but each points at a different layer of the same ecosystem.
This article maps out Anthropic as a company, the research pedigree behind it, and the products that currently exist under the Claude name as of 2026. It's written for people evaluating Claude for work, or who find themselves fuzzy on the difference between, say, Claude Pro and the Claude API.
Anthropic and Claude Labs Are Not the Same Thing
A quick terminology pass first.
Anthropic is an American AI research company founded in 2021, headquartered in San Francisco. Its stated focus is AI safety, and much of its research comes out of that lens. Co-founders Dario and Daniela Amodei brought in many former OpenAI researchers, including several who were instrumental in GPT-3.
Claude is the family of large language models that Anthropic builds. It's not one model — it's a lineup that evolves in generations. As of 2026, the active trio is Opus 4.6, Sonnet 4.6, and Haiku 4.5.
Claude Labs isn't an official entity. It's an informal way people refer to Anthropic's experimental or research-flavored work, and it's sometimes used loosely for Anthropic as a whole. Our own site, Claude Lab, borrows the name in the spirit of "a lab for using Claude well" — we're independent from the company and publish from a practitioner's perspective.
So the hierarchy is: Anthropic is the company. Claude is the model family. "Claude Labs" is colloquial.
Why Safety Is Anthropic's Calling Card
If you've used Claude alongside other models, you may have noticed its responses tend to be more measured, and its refusals tend to explain themselves rather than shutting the door. That isn't an accident. It's the company's positioning made tangible.
Many of the founders came from OpenAI, motivated by building what they saw as safer AI from an independent vantage point. In 2022, Anthropic published the Constitutional AI method — a training approach where the model is taught to critique and revise its own outputs against stated principles. The "thoughtful but grounded" feel of modern Claude is a direct descendant of that work.
On the funding side, Google and Amazon have together invested several billion dollars, which is why Claude can be accessed via AWS Bedrock and Google Vertex AI in addition to Anthropic's own platform. That unusual distribution setup is part of what makes "where do I consume Claude from?" a real question below.
The Four Product Lines to Know About
As of 2026, the products you'll encounter as a user or builder cluster into four groups.
1. Claude.ai (the Web App)
The most recognizable surface is the browser app at Claude.ai. There's a free tier, and paid tiers that step up in frequency and model access: Pro (individual), Max (heavy personal use), Team (small groups), and Enterprise.
The free tier includes access to the Sonnet family, which makes it a great zero-friction starting point. Upgrading unlocks more messages per day and higher-end models.
2. The Claude API (Anthropic Platform)
If you're building software that calls Claude, you sign up on the Anthropic Platform and get an API key. You pay per input/output token, with rates varying per model.
AWS Bedrock and Google Vertex AI expose the same Claude models, with slightly different key-issuing flows and version availability. The rule of thumb: AWS-centric organizations go through Bedrock, Google Cloud shops through Vertex AI, and everyone else goes directly through Anthropic.
3. Claude Code
A terminal-based agentic coding tool. Claude reads and writes files, runs shell commands, and drives git, all autonomously. It has been a major driver of productivity gains among solo developers since its late-2024 debut.
We cover Claude Code heavily on this site. A good starting point is our complete reference for Claude Code environment variables.
4. Claude Agent SDK / Claude Platform
For developers who want to embed agentic behavior into their own products, the Claude Agent SDK packages up the tool-calling, conversation loop, and memory management that power Claude Code. Alongside it, the broader Claude Platform is maturing — with Skills (reusable instruction bundles for agents) and MCP (Model Context Protocol) for wiring up external tools in a standard way.
How to Pick a Tier
The right entry point depends on who you are and what you're building.
- Daily personal use: start on Claude.ai's free tier. Move to Pro if you hit the cap, Max if you use it all day
- Coding as an individual developer: install Claude Code and pay per-token via an Anthropic API key
- Small team of ~10–50 people: a Team plan shared workspace, plus API keys for anyone building something
- Production integration: Bedrock if you're on AWS, Vertex if you're on Google Cloud, Anthropic Platform otherwise
- Enterprise: the Enterprise plan, which brings SSO, audit logs, and a dedicated point of contact
Exact pricing shifts more often than I can keep up with, so always check the official pricing page before making a decision. The list above is about who fits where, not what the current numbers are.
A Sensible Learning Path
"Claude is smart" is everywhere; what's missing is a map of which product to use when. Start on Claude.ai to feel how Claude behaves. Once you're comfortable, try Claude Code on a real project. Come to the API only when your work starts leaving the terminal and going into software you build.
For the Claude Code side, our complete guide to settings.json is worth bookmarking. For API-first work, the api-sdk category on this site has deeper material. Follow your goals, not the category order.