llm-selector

Choose the right LLM, then verify you're actually using it.

llm-selector is a lightweight, zero-dependency advisor for small and medium-sized businesses choosing between GLM, Qwen, DeepSeek, Kimi, Claude, GPT, and local open-source models.

It is not another benchmark leaderboard. It starts from the customer's business scenario, budget, data sensitivity, and deployment preference, then generates a practical shortlist and next-step report.

Why this exists

Most LLM benchmarks answer: "Which model scores higher on a dataset?"

SMEs usually ask something more practical:

• Which model should we use for AI coding, documents, customer service, or internal knowledge?
• Should we use hosted APIs, domestic providers, hybrid routing, or private deployment?
• What risks should we review before sending company data to a model?
• If a gateway claims to route to GLM-5.2, how do we verify it?
• What should our first PoC look like?

llm-selector turns those questions into a simple selection report.

Who is this for?

• SME founders, OPCs, freelancers, and IT managers evaluating LLM adoption
• Developers helping a company or individual choose an AI stack
• Teams comparing GLM, Qwen, DeepSeek, Kimi, Claude, GPT, and local models
• Consultants preparing an initial AI model selection report
• Anyone who needs practical model-selection guidance before building a PoC

What it does

The CLI asks about:

• business scenario
• data sensitivity
• budget preference
• deployment preference

Then it outputs:

• top 3 model/model-family recommendations
• why each option fits
• risks to review
• verification advice
• suggested next steps for PoC

Usage

git clone https://github.com/xiaoyanfei-tech/llm-selector.git
cd llm-selector
python selector.py

Non-interactive example:

python selector.py \
  --scenario ai_coding \
  --scenario document_summary \
  --sensitivity internal \
  --budget medium \
  --deployment domestic \
  --output report.md

Machine-readable output:

python selector.py --scenario ai_coding --json

Share or feedback templates:

python selector.py --share-template
python selector.py --feedback-template

Example output

Wrote report.md
Top recommendations:
- GLM-5.2 (10)
- Qwen (7)
- DeepSeek (6)

The generated Markdown report includes:

# SME LLM Selection Report

> Choose the right LLM, then verify you're actually using it.

## Recommended shortlist

### 1. GLM-5.2
Why it fits:
- Strong fit for Claude Code-style AI coding
- Long-context friendly

Risks to review:
- Verify the endpoint is really GLM-5.2 before trusting a gateway

Verification: Run verify-glm...

Relationship with verify-glm

llm-selector helps choose the right model direction.

verify-glm verifies whether a GLM-5.2 endpoint is actually behaving like GLM-5.2.

Together:

Choose the right LLM → verify you're actually using it → run a focused PoC

Supported scenarios

• AI coding / developer assistant
• Enterprise knowledge base / RAG
• Customer service chatbot
• Document summary / contract reading
• Translation / bilingual writing
• Data analysis / report assistant
• Office automation / internal assistant

Supported model families

• GLM-5.2
• Qwen
• DeepSeek
• Kimi
• Claude
• GPT family
• Local open-source models

Commercial-ready next steps

This repo can be used as a lead-capture and consulting workflow:

• Start with the CLI selection report
• Read ROADMAP.md to see where the project is going
• Read CONTRIBUTING.md to contribute scenarios, cases, or model notes
• Read docs/vision.md for the long-term product thesis
• Read docs/go-to-market.md for the market-first strategy
• Read docs/market-validation.md to track evidence and traction
• Read docs/case-studies.md for free consultation case studies based on real public questions
• Read docs/hot-topics/ for daily hot topics with free solution directions
• Read docs/methodology.md to understand the decision framework
• Use docs/client-intake.md to collect business context
• Use docs/packages.md to choose the smallest useful next step
• Use docs/report-template.md for a paid selection report
• Use docs/poc-plan.md for a 2-4 week implementation plan
• Use docs/services.md to describe available consulting deliverables

Potential paid deliverables:

• OPC / personal LLM selection
• LLM selection report
• AI coding stack recommendation
• endpoint verification report
• PoC plan and integration checklist

中文说明

中文用户请看 README.zh-CN.md。

Need help?

Need help choosing an LLM stack?

Open a selection request:

https://github.com/xiaoyanfei-tech/llm-selector/issues/new/choose

If you are choosing an LLM stack for yourself, an OPC, a small team, or a small/medium-sized business, this tool can produce a first-pass report. For deeper help, open an issue with your use case:

• business scenario
• data sensitivity
• estimated monthly usage
• preferred providers or deployment constraints
• current tools such as Claude Code, Cursor, Dify, FastGPT, or internal systems

Possible consulting deliverables:

• OPC / personal LLM selection
• LLM selection report
• AI coding stack recommendation
• enterprise knowledge-base model recommendation
• endpoint verification report
• PoC plan and integration checklist

Limitations

• This is a practical decision aid, not a universal benchmark.
• Model quality, price, and availability change quickly.
• Always validate recommendations with your own tasks and data policy.
• The included model data is intentionally simple and should be updated as the market changes.

License

MIT — see LICENSE.