Skip to the content.

Vision

AI model selection is becoming a business decision layer.

Benchmarks are useful, but they do not answer the full customer question:

Which model should I use for my workflow, budget, data policy, and deployment constraints?

llm-selector starts from the customer side instead of the benchmark side.

Thesis

The number of LLM options will keep growing:

For OPCs, small teams, and SMEs, the problem is not lack of options. The problem is decision risk.

They need:

Product wedge

The first wedge is a free open-source CLI that generates practical selection reports.

The second wedge is verification:

llm-selector chooses
verify-glm verifies

The long-term opportunity is a decision platform for AI adoption:

scenario intake → model shortlist → verification → PoC plan → case feedback → better recommendations

Why this can become defensible

The defensible asset is not the initial code. It is the compounding case library:

A public benchmark says which model scores higher. A case library says which stack worked for which business constraint.

Market posture

Win trust before monetizing heavily.

This keeps adoption easy while preserving a path to consulting, reports, and future platform revenue.