llm LLM
Large Language Model
A neural network trained on vast amounts of text data that can generate, understand, and analyze natural language text.
What Is an LLM
A Large Language Model (LLM) is a type of AI built on the transformer architecture. The model is trained on billions of text documents and can:
- Generate text based on instructions (prompts)
- Answer questions
- Translate between languages
- Write and analyze code
- Act as an agent with tools
Popular LLMs
- Claude Opus 4 (Anthropic) — flagship reasoning model
- GPT-5 (OpenAI) — frontier general model
- Gemini 2.5 Pro (Google DeepMind) — 1M+ context, multimodal
- Llama 4 (Meta) — open-weights flagship
- Mistral Large (Mistral AI) — European, open-friendly
- Grok 4 (xAI) — frontier reasoning, real-time access
Other frontier-lab players: Perplexity builds a search-first LLM product on top of frontier models; Cohere targets the enterprise RAG market; Hugging Face hosts the model weights ecosystem most open releases ship through; Scale AI provides the training-data labeling infrastructure behind several frontier labs.
Key Parameters
- Context window — how much text the model can process at once
- Temperature — level of “creativity” in responses (0 = deterministic, 1 = creative)
- Tokens — units of text the model works with (~4 characters = 1 token in English)