LLM (large language model)
An AI model trained on huge amounts of text that can write, summarise, classify and answer questions, offered by providers such as OpenAI, Anthropic and Google.
Updated October 2026 · with examples from our own builds
What does LLM mean in practice?
Large language models power chatbots, writing tools, form builders and assistants. Your product sends text, sometimes with images, and receives text back, paying per unit of text processed. Different models trade quality, speed and price against each other.
A useful mental model is a fluent assistant that doesn't know your business and sounds confident either way. LLMs are good at phrasing and pattern-matching and unreliable as a source of facts, so products that use them well let the model write while code and data decide what is true.
Keeping the provider open is easier than it seems. If the model is a configuration setting rather than built into every feature, switching later is a change and a test run, not a rebuild.
LLM at a glance
- Good at
- Writing, summarising, classifying and planning
- Not reliable as
- A source of facts about your business
- Keep swappable
- The provider and model, as configuration
Where has LLM come up in our builds?
131 specialist copilots across 20 domains run on hosted AI models, each configured for one kind of problem so the user describes their situation instead of engineering a prompt.
The whole AI exchange runs at temperature 0 with JSON output, a 60-second deadline and user messages capped at 2,000 characters, and which provider runs it is a configuration setting, not a rebuild.
The model provider is a setting on each bot, with 21 providers listed, so switching models doesn't mean rebuilding the conversation flow.
Common questions about LLM
Can I switch LLM providers after launch?
Yes, if the provider and model are configuration rather than code spread across features. AskForth and ConferBot both treat the provider as a setting, so a switch is a change plus a test run, not a rebuild.
Do I need to train my own AI model?
Rarely. Most products get further by configuring an existing model well, grounding it in their own data and checking its output in code than by training one.
Related terms
Planning a product and not sure where LLM fits? Ask us, or browse all 116 terms in the glossary.