What is Large Language Model (LLM)?
A neural network trained on massive text corpora to predict tokens, enabling generation, reasoning, and tool use via natural language.
Definition
Large language models are transformer-based networks—GPT, Claude, Gemini, Llama—pretrained on trillions of tokens to predict the next token, then aligned with instruction tuning and reinforcement learning from human feedback. Emergent capabilities include code synthesis, summarization, translation, and multi-step tool calling, accessed through chat or API with a bounded context window measured in tokens. Key operational realities: outputs are probabilistic (temperature controls variance), hallucination requires grounding via retrieval, and cost scales with input plus output tokens. Developers integrate LLMs through provider APIs or self-hosted open-weight models served by vLLM or Ollama.