An LLM is a neural network trained on huge amounts of text (and often other media) to predict the next token in a sequence — a simple training objective that, at scale, produces systems capable of conversation, reasoning, coding, and summarization. "LLM" is the umbrella tag covering GPT, Claude, Gemini, Llama, DeepSeek, Mistral and every other model in this class, which is why it ranks #2 in trending tech usage at roughly 27.6% share — it's the term underlying almost every other AI tag on this list. There's no single vendor; the concept applies across dozens of competing model families and every major programming platform.
Every LLM interaction follows the same basic pipeline: your text is tokenized, fed through the model, and the generated tokens are turned back into text. Most hosted LLMs expose this as a "chat completion" API — this example uses the OpenAI-compatible format supported by many providers.
# pipeline: prompt -> tokenize -> model forward pass -> sample tokens -> detokenize -> text
import requests
response = requests.post(
"https://api.openai.com/v1/chat/completions",
headers={"Authorization": "Bearer YOUR_KEY"},
json={
"model": "gpt-4o-mini",
"messages": [
{"role": "user", "content": "Summarize the plot of Hamlet in one sentence."}
]
}
)
print(response.json()["choices"][0]["message"]["content"])The fastest path is a hosted API — sign up with any provider, grab a key, and send your first chat request. To run a model with zero API cost, use a local runtime like Ollama instead.
# hosted route: pick a provider, get an API key, then call its chat endpoint
# e.g. OpenAI, Anthropic, Google AI Studio, or Mistral La Plateforme all offer free-tier keys
# local route: no API key needed
ollama run llama3