Three layers explain most of the confusion
A company develops or distributes technology. A model is a trained system with a particular version and configuration. An app gives you an interface, files, history, permissions, and tools. The same model can appear in more than one app.
OpenAI develops GPT models and offers ChatGPT. Anthropic develops Claude models and offers the Claude app. Google develops Gemini models and the Gemini app, as well as the distinct Gemma open-model family. Model names and app names sometimes overlap; that does not make their features identical.
Qwen is a family of models from Alibaba’s Qwen team. Some Qwen releases can be downloaded and run in compatible local apps, including LM Studio. A local app supplies the interface and runtime; it does not turn every model it loads into the same product.
Why the experience can differ
An app may browse the web, retrieve files, use a calculator, or select a model automatically. Its instructions and account settings also affect the response. Two screenshots labeled with the same model name may therefore represent different setups.
Choose around the job
For a first experiment, choose an app you can access comfortably and a task with an easily checked result. For local use, model size, available memory, supported software, and the exact download matter. The biggest model on a leaderboard may be impractical for your computer.
Check the model picker and the provider’s current documentation before assuming a feature is included in a plan. Our model field guide is a dated snapshot with source links, not a promise that every account sees every model.
A useful comparison
Try the same short task in two tools, with fresh chats and identical input. Note which app and version you used, whether browsing was enabled, and how much correction the result required. You are comparing complete experiences unless you have controlled the surrounding software too.