Glossary
LLM
An LLM, or Large Language Model, is an artificial intelligence specialized in processing human language.
An LLM is a computer model trained on billions of words to understand, generate, and rephrase text in natural language.
In other words, an LLM is like a super digital brain that can:
- answer questions,
- write texts (emails, articles, summaries...),
- translate languages,
- write code,
- hold fluid conversations...
These models work using a technology called neural networks (especially transformers) and learn by analyzing vast amounts of data (books, websites, articles, etc.).
🧠What is an LLM used for?
- Personal or professional assistant: note-taking, summarizing, content generation.
- Automated customer support.
- Developer tools (code generation or review).
- Creative writing (stories, scripts, poems…).
- Translation, summarization, classification, semantic search, and more.
🆚 Comparison of major LLMs in 2025
| Criterion | OpenAI (ChatGPT) | Anthropic (Claude) | Google (Gemini) | Mistral |
|---|---|---|---|---|
| Company | OpenAI | Anthropic | Google DeepMind | Mistral AI (French startup 🇫🇷) |
| Flagship Model (2025) | GPT-4o | Claude 3 (Opus, Sonnet...) | Gemini 1.5 Pro | Mistral Large / Mistral 7B |
| Strengths | Multimodal, high overall performance, strong ecosystem (plugins, API) | Long context, structured reasoning, safe outputs | Strong in math and research | Open source, fast, lightweight and efficient |
| Context Window | ~128k tokens (GPT-4o) | up to 200k tokens | up to 1 million tokens | more limited, depending on the model |
| Multimodality | Text, image, audio, video | Text only (image coming soon) | Text, image, code | Text only (as of now) |
| Ethics / Safety Focus | Well balanced | Strong focus on safety, neutrality, refusal handling | Controlled by Google, strict filtering | Maximum freedom, lightly filtered |
| Access | ChatGPT, API, Microsoft Copilot | Claude.ai, API via Slack, Notion, etc. | Gemini via Bard, Workspace, Android | API, free integration, open source |
| License | Proprietary | Proprietary | Proprietary | Open source (for smaller models) |
| Primary Language | English, strong multilingual support | English (improving in French) | Multilingual | Very strong in French |
| Typical Use | General public, businesses, developers | Businesses, writers, legal professionals | Research, coding, Android users | Developers, startups, open source communities |
🧩 Key differences summary
- OpenAI (ChatGPT): Versatile, suited for both general and professional users, highly integrated with the Microsoft ecosystem. Great all-around, but closed-source.
- Claude (Anthropic): Focused on reasoning and safety, excellent for long documents, and more cautious in its answers.
- Gemini (Google): Part of the Google ecosystem, excellent at math, research, and multimodal input. Can feel more "robotic."
- Mistral: Open source, fast, European, great for tech projects, but lacks user-friendly interfaces and is not yet multimodal.
Let me know if you'd like a version tailored for a specific audience (e.g. developers, business users, educators).