Large Language Model (LLM)

What is a large language model?

A large language model is an AI model trained on very large volumes of text to predict what comes next in a sequence. That single mechanism, applied at sufficient scale, produces the ability to answer questions, summarise, translate, write code, follow instructions, and reason through multi-step problems.

The important consequence of the mechanism is that an LLM has no internal database and no lookup step. It does not retrieve a fact and report it; it generates the most probable continuation given its input. This explains both its flexibility and its characteristic failure mode, since a plausible continuation and a true one are usually the same and not always.

Modern LLMs are also instruction tuned, meaning they have been further trained to follow directions and produce helpful responses rather than simply continuing text. This is what makes them usable as components rather than as curiosities.

What determines which LLM an enterprise should use?

Public benchmark rankings are a weak guide, because they measure general capability while enterprises run specific workloads. Four criteria matter more.

Measured performance on your own cases. An evaluation set built from real historical work with known correct outcomes. Models that lead public leaderboards routinely underperform smaller ones on narrow enterprise tasks.

Cost at production volume. Cost per resolved case rather than per token. A stronger model that succeeds in one pass can be cheaper than a weaker one that retries and escalates.

Latency against the interaction pattern. Voice and interactive assistance impose response constraints that batch document processing does not.

Deployment and sovereignty constraints. Where regulation or contract restricts processing location, the eligible set narrows, sometimes to open weight models in a controlled environment. This frequently overrides the other three.

Most mature deployments end up using several models, with cheaper ones handling high volume classification and extraction and stronger ones reserved for reasoning heavy steps.

How Wizr AI selects language models per engagement

Wizr AI describes a model agnostic architecture spanning proprietary and open source models, with the stack selected per engagement based on performance, cost, and data sovereignty requirements rather than on a fixed vendor relationship.

This is an architectural position rather than a preference. Because retrieval, tool definitions, orchestration, and the security and governance controls sit in the platform rather than in the model, substituting the reasoning layer does not require rebuilding the integrations or renegotiating the controls.

Where a model needs adaptation rather than replacement, fine tuning of LLMs and AI models is offered within Enterprise AI Services, alongside AI Ops for continuous management and optimisation of data and models after deployment.