The AI Market: Who Does What?
iAs of: September 2026
The information in this section reflects the state as of September 2026. Model names, prices, and market data change frequently — some details may have already changed.
Wissen
The "Big Players" of the AI World
Think of the AI market as a competition between technology giants where new records are broken every week. As of September 2026, six major companies dominate the field:
OpenAI (San Francisco) -- The pioneer of modern AI. Known for ChatGPT, which has over 900 million weekly active users (as of February 2026), making it the most-used AI application in the world. Their current top model is GPT-5.6 (July 2026). OpenAI is financially backed by Microsoft and was the first company to make AI accessible to the masses.
Anthropic (San Francisco) -- Founded by former OpenAI employees with a focus on AI safety. Their premium model Claude Fable 5 (June 2026) comes from the Mythos class; the strongest all-round model is Claude Opus 5 (July 2026), which leads the coding benchmarks. Anthropic pursues a "safety first" approach -- AI should be reliable and honest.
Google DeepMind (London/Mountain View) -- The merger of Google's AI labs. Gemini 3.1 Pro (February 2026) is deeply integrated into Google's products (Search, Gmail, YouTube, Maps). Google has the advantage of building AI directly into products that billions of people use daily.
Meta (Menlo Park) -- Mark Zuckerberg's company is fully committed to open source. Llama 4 can be downloaded and used by anyone -- even commercially. Meta wants AI to be accessible to everyone, not just paying customers.
xAI (San Francisco) -- Elon Musk's AI company with the model Grok. Closely linked to the platform X (formerly Twitter) and known for a casual, uncensored communication style.
Mistral (Paris) -- Europe's AI hope. The French startup shows that innovative AI does not have to come only from Silicon Valley. Mistral focuses on efficient, high-performance models and is an important player in the open-source space.
iDid You Know?
OpenAI was founded in 2015 as a nonprofit organization -- with the goal of developing AI for the benefit of humanity. It has since become one of the most valuable technology companies in the world. Anthropic was founded in 2021 by former OpenAI employees who wanted more focus on AI safety.
Open Source vs. Closed Source Models
In the AI world, there is an important distinction that divides the industry:
Closed source: The program code and model weights are secret. You can only use the model through the provider's website or API (Application Programming Interface). Examples: GPT-5.6, Claude Opus 5, Gemini 3.1 Pro.
Open source: The code and model weights are publicly available. Anyone can download the model, customize it, and even create their own versions. Examples: Llama 4 (Meta), Mistral models.
| Closed Source | Open Source | |
|---|---|---|
| Advantage | Often more powerful, easier to use | Free, customizable, transparent |
| Disadvantage | Cost, dependency on the provider | Requires technical knowledge, own hardware |
| Examples | GPT-5.6, Claude, Gemini | Llama 4, Mistral, Qwen |
What is the main advantage of open-source AI models?
Match the AI models to the correct companies:
Current Models 2026: Context Sizes Compared
An important measure for AI models is the context size -- that is, how much text the model can "read" and process at once. This is measured in tokens (a token corresponds roughly to a word or word part).
| Model | Context Size | Roughly Equivalent To |
|---|---|---|
| GPT-5.6 | 1M Tokens | ~750,000 words (about 12 books) |
| Claude Opus 5 | 1M Tokens | ~750,000 words (about 12 books) |
| Gemini 3.1 Pro | 1M Tokens | ~750,000 words (about 12 books) |
| Llama 4 Scout | 10M Tokens | ~7,500,000 words (about 125 books) |
128K Tokens
GPT-4 Turbo / GPT-4o
Books
~1.4 books
Pages
~191 Pages
Code Files
~512
~250 Tokens each
Chat Messages
~3,200
~40 Tokens each
Context Window Comparison
Note: Values are approximations. 1 Token equals roughly 3/4 of an English word.
*What Does That Mean in Practice?
With 1 million tokens of context (e.g. Claude Opus 5, GPT-5.6, or Gemini 3.1 Pro), you can give the model an entire book or hundreds of code files and ask questions about them. Llama 4 Scout with 10 million tokens can theoretically process an entire library -- though only on sufficiently powerful hardware.
Verstehen
Think of the difference like recipes: closed source is like visiting a restaurant -- you get the finished dish but do not know the recipe. Open source is like a cookbook -- you have the complete recipe and can change it to your taste.
Cost Overview: How Much Does AI Cost?
For end users, there are usually free starter versions. If you need more power, you pay a monthly subscription:
| Service | Free | Subscription Price (approx.) | What You Get |
|---|---|---|---|
| ChatGPT | Yes (GPT-5 mini) | ~20 USD/month (Plus) | GPT-5.6, more messages, image generation |
| Claude | Yes (Sonnet) | ~20 USD/month (Pro) | Opus 5, more messages, extended features |
| Gemini | Yes (in Google products) | ~20 USD/month (Advanced) | Gemini 3.1 Pro, 1M context, Notebook LM |
For developers who want to integrate AI into their own applications, there is API access (Application Programming Interface). Here you pay per processed token:
- Claude Fable 5 API: $10 (input) / $50 (output) per 1 million tokens — the new top model
- GPT-5.6 API: $4 (input) / $20 (output) per 1 million tokens (Sol)
- Claude Opus 5 API: $5 (input) / $25 (output) per 1 million tokens
- Gemini 3.1 Pro API: $2 (input) / $12 (output) per 1 million tokens — the cheapest provider
- Mistral Large 3 API: $0.50 (input) / $1.50 (output) per 1 million tokens — extremely affordable
- Llama 4: Free to download, but own hardware required
Which AI model has the largest context size as of September 2026?
Anwenden
There is no "one best model" -- the choice depends on your specific use case. To get started, try the free versions of ChatGPT and Claude and compare them. This helps you develop a feel for which tool works best for which task.
Reflect
The AI market is developing rapidly. New models appear on a monthly basis, prices are falling, and performance is rising. For you as a beginner, it is important to know: there is no "one best model." Each has strengths and weaknesses, and the choice depends on your specific use case. In the next module, you will learn about the most important AI tools in detail and find out how to use them in everyday life.