LLM Checker 2.0 🚀

Advanced CLI tool that scans your hardware and tells you exactly which LLM or sLLM models you can run locally, with full Ollama integration.

✨ What’s New in v2.0

• 🦙 Full Ollama integration – Detects installed models, benchmarks performance and handles downloads automatically
• 🐣 sLLM (Small Language Model) support – From 0.5 B all the way up to ultra‑efficient models
• 📊 Expanded model database – 40 + models including Gemma 3, Phi‑4, DeepSeek‑R1, Qwen 2.5
• 🎯 Improved compatibility analysis – Granular 0‑100 scoring system
• 🏷️ Detailed categorisation – ultra‑small, small, medium, large, embedding, multimodal
• ⚡ Performance estimation – tokens/s, memory footprint, energy consumption
• 🧠 Use‑case‑based recommendations – general, code, chat, embeddings, multimodal
• 📱 Redesigned CLI – cleaner UX with colours & emojis

🚀 Installation

Option 1 – Global NPM (recommended)

npm install -g llm-checker

Option 2 – With Ollama (recommended for running models)

# 1 Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# 2 Install LLM Checker
npm install -g llm-checker

# 3 Verify
llm-checker check

Option 3 – Local development

git clone https://github.com/developer31f/llm-checker.git
cd llm-checker
npm install
npm link

📖 Usage

Main command – full analysis

# Full system scan + Ollama detection
llm-checker check

# Detailed hardware info
llm-checker check --detailed

# Run a performance benchmark
llm-checker check --performance-test

Filter by model category

llm-checker check --filter ultra_small   # Models < 1 B params
llm-checker check --filter small         # 1–4 B
llm-checker check --filter medium        # 5–15 B
llm-checker check --filter large         # > 15 B

Filter by specialisation

llm-checker check --filter code         # Programming models
llm-checker check --filter chat         # Conversational
llm-checker check --filter multimodal   # Vision + text
llm-checker check --filter embeddings   # Embedding models

Use‑case presets

llm-checker check --use-case code        # Optimised for coding
llm-checker check --use-case chat        # Optimised for conversation
llm-checker check --use-case embeddings  # Semantic search

Ollama‑only / include cloud

llm-checker check --ollama-only          # Only local models
llm-checker check --include-cloud        # Include cloud models for comparison

Ollama management

# List installed models
llm-checker ollama --list

# Show running models
llm-checker ollama --running

# Benchmark a specific model
llm-checker ollama --test llama3.2:3b

# Download a model
llm-checker ollama --pull phi3:mini

# Remove a model
llm-checker ollama --remove old-model:tag

Explore the model database

llm-checker browse                 # All models
llm-checker browse --category small
llm-checker browse --year 2024
llm-checker browse --multimodal

Help commands

llm-checker --help                # Global help
llm-checker check --help          # Help for a specific command
llm-checker ollama --help

🎯 Example output

High‑end system with Ollama

🖥️  System Information:
CPU: Apple M2 Pro (12 cores, 3.5 GHz)
Architecture: Apple Silicon
RAM: 32 GB total (24 GB free, 25 % used)
GPU: Apple M2 Pro (16 GB VRAM, dedicated)
OS: macOS Sonoma 14.2.1 (arm64)

🏆 Hardware Tier: HIGH (Overall Score: 92/100)

🦙 Ollama Status: ✅ Running (v0.1.17)
📦 Local Models: 5 installed
🚀 Running Models: llama3.1:8b

⚡ Performance Benchmark:
CPU Score: 95/100
Memory Score: 88/100
Overall Score: 91/100

✅ Compatible Models (Score ≥ 75):
┌─────────────────────┬──────────┬───────────┬──────────┬──────────┬───────────┬──────────┐
│ Model               │ Size     │ Score     │ RAM      │ VRAM     │ Speed     │ Status   │
├─────────────────────┼──────────┼───────────┼──────────┼──────────┼───────────┼──────────┤
│ Llama 3.1 8B        │ 8 B      │ 98/100    │ 8 GB     │ 4 GB     │ medium    │ 📦 🚀    │
│ Mistral 7B v0.3     │ 7 B      │ 97/100    │ 8 GB     │ 4 GB     │ medium    │ 📦       │
│ CodeLlama 7B        │ 7 B      │ 97/100    │ 8 GB     │ 4 GB     │ medium    │ 💻       │
│ Phi‑3 Mini 3.8B     │ 3.8 B    │ 99/100    │ 4 GB     │ 2 GB     │ fast      │          │
│ Gemma 3 1B          │ 1 B      │ 100/100   │ 2 GB     │ 0 GB     │ very_fast │          │
└─────────────────────┴──────────┴───────────┴──────────┴──────────┴───────────┴──────────┘

Resource‑limited system

🖥️  System Information:
CPU: Intel Core i5‑8400 (6 cores, 2.8 GHz)
Architecture: x86‑64
RAM: 8 GB total (3 GB free, 62 % used)
GPU: Intel UHD Graphics 630 (0 GB VRAM, integrated)
OS: Ubuntu 22.04 LTS (x64)

🏆 Hardware Tier: LOW (Overall Score: 45/100)

🦙 Ollama Status: ❌ Ollama not running (connection refused)

🔧 Supported models (40 +)

🐣 Ultra‑small (< 1 B params)

• Qwen 0.5B – Ultra lightweight, requires 1 GB RAM
• LaMini‑GPT 774 M – Multilingual compact model, 1.5 GB RAM

🐤 Small (1 – 4 B)

• TinyLlama 1.1 B – Perfect for testing, 2 GB RAM
• Gemma 3 1 B – Mobile‑optimised, 2 GB RAM, 32 K context
• MobileLLaMA 1.4 B / 2.7 B – 40 % faster than TinyLlama
• Llama 3.2 1 B / 3 B – Compact Meta models
• Phi‑3 Mini 3.8 B – Great reasoning from Microsoft, 4 GB RAM
• Gemma 2 B – Efficient Google model, 3 GB RAM

🐦 Medium (5 – 15 B)

• Llama 3.1 8 B – Perfect balance, 8 GB RAM
• Mistral 7 B v0.3 – High‑quality EU model, 8 GB RAM
• Qwen 2.5 7 B – Multilingual with strong coding ability
• CodeLlama 7 B – Specialised for coding, 8 GB RAM
• DeepSeek Coder 6.7 B – Advanced code generation
• Phi‑4 14 B – Latest Microsoft model with improved capabilities
• Gemma 3 4 B – Multimodal with long context (128 K)

🦅 Large (> 15 B)

• Llama 3.3 70 B – Meta flagship, 48 GB RAM
• DeepSeek‑R1 70 B – Advanced reasoning (o1‑style)
• Mistral Small 3.1 22 B – High‑end EU model
• Gemma 3 12 B / 27 B – Google multimodal flagships
• CodeLlama 34 B – Heavy coding tasks, 24 GB RAM
• Mixtral 8×7 B – Mixture‑of‑Experts, 32 GB RAM

🖼️ Multimodal (Vision + text)

• LLaVA 7 B – Image understanding, 10 GB RAM
• LLaVA‑NeXT 34 B – Advanced vision capabilities
• Gemma 3 4 B / 12 B / 27 B – Google multimodal family

🧲 Embedding models (semantic search)

• all‑MiniLM‑L6‑v2 – Compact 0.5 GB embedding model
• BGE‑small‑en‑v1.5 – High‑quality English embeddings

☁️ Cloud models (for comparison)

• GPT‑4 – OpenAI, requires API key & internet
• Claude 3.5 Sonnet – Anthropic, 200 K context

🛠️ Advanced Ollama integration

Automatic model management

llm-checker ollama --list       # Details of every local model
llm-checker ollama --running    # Monitor VRAM usage
llm-checker ollama --test llama3.1:8b

Smart installation

# After analysis, install all recommended models automatically
llm-checker check --filter small | grep "ollama pull" | bash

Real‑time model comparison

for model in $(ollama list | grep -v NAME | awk '{print $1}'); do
  echo "Testing $model:"
  llm-checker ollama --test $model
done

📊 Detailed compatibility system

Scoring scale (0‑100)

Compatibility factors

1. Total RAM vs requirement (40 %)
2. Available VRAM (25 %)
3. CPU cores (15 %)
4. CPU architecture (10 %)
5. Quantisation availability (10 %)

Hardware tiers

• 🚀 ULTRA_HIGH – 64 GB RAM, 32 GB VRAM, 12+ cores
• ⚡ HIGH – 32 GB RAM, 16 GB VRAM, 8+ cores
• 🎯 MEDIUM – 16 GB RAM, 8 GB VRAM, 6+ cores
• 💻 LOW – 8 GB RAM, 2 GB VRAM, 4+ cores
• 📱 ULTRA_LOW – below the above

⚙️ Advanced configuration

Environment variables

export OLLAMA_BASE_URL=http://localhost:11434
export LLM_CHECKER_RAM_GB=16
export LLM_CHECKER_VRAM_GB=8
export LLM_CHECKER_CPU_CORES=8
export LLM_CHECKER_LOG_LEVEL=debug
export LLM_CHECKER_CACHE_DIR=/custom/cache
export NO_COLOR=1

Configuration file ~/.llm-checker.json

{
  "analysis": {
    "defaultUseCase": "code",
    "performanceTesting": true,
    "includeCloudModels": false
  },
  "ollama": {
    "baseURL": "http://localhost:11434",
    "enabled": true,
    "timeout": 30000
  },
  "display": {
    "maxModelsPerTable": 15,
    "showEmojis": true,
    "compactMode": false
  },
  "filters": {
    "minCompatibilityScore": 70,
    "excludeModels": ["very-large-model"]
  },
  "customModels": [
    {
      "name": "Custom Model 7B",
      "size": "7B",
      "requirements": { "ram": 8, "vram": 4 }
    }
  ]
}

🎮 Specific use cases

Coding

llm-checker check --use-case code --filter medium
ollama pull $(llm-checker check --use-case code --ollama-only | grep "ollama pull" | head -1 | awk '{print $3}')
echo "def fibonacci(n):" | ollama run codellama:7b "Complete this Python function"

Chatbots / assistants

llm-checker check --use-case chat --filter small,medium
ollama pull llama3.2:3b
ollama run llama3.2:3b "Hello! How can you help me today?"

Semantic search / RAG

llm-checker check --filter embeddings
ollama pull all-minilm

Multimodal analysis

llm-checker check --multimodal
ollama pull llava:7b
# echo "image.jpg" | ollama run llava:7b "Describe this image"

🔍 Troubleshooting

Ollama not detected

curl http://localhost:11434/api/version          # Should return JSON

# Install or start service
curl -fsSL https://ollama.ai/install.sh | sh
sudo systemctl start ollama                      # Linux

Incorrect hardware detection

llm-checker check --detailed
export LLM_CHECKER_RAM_GB=16
export LLM_CHECKER_VRAM_GB=8
llm-checker check

Models marked as incompatible

llm-checker check --min-score 0
llm-checker check --include-marginal
llm-checker check --quantization Q2_K

Permission errors (Linux/macOS)

sudo chmod +r /sys/class/dmi/id/*
sudo chmod +r /proc/meminfo
sudo llm-checker check

🧪 Programmatic API (Node)

const LLMChecker = require('llm-checker');
const OllamaClient = require('llm-checker/ollama');

const checker = new LLMChecker();
const analysis = await checker.analyze({
  useCase: 'code',
  includeCloud: false,
  performanceTest: true
});

console.log('Compatible models:', analysis.compatible);

const ollama = new OllamaClient();
const localModels = await ollama.getLocalModels();

🚀 CI/CD Workflows

GitHub Actions

name: LLM Compatibility Check
on: [push, pull_request]

jobs:
  llm-check:
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v4

    - name: Setup Node.js
      uses: actions/setup-node@v4
      with:
        node-version: '18'

    - name: Install LLM Checker
      run: npm install -g llm-checker

    - name: Check LLM Compatibility
      run: |
        llm-checker check --format json > compatibility.json
        cat compatibility.json

    - name: Upload Results
      uses: actions/upload-artifact@v4
      with:
        name: llm-compatibility
        path: compatibility.json

Docker integration

FROM node:18-alpine

RUN apk add --no-cache dmidecode lm-sensors
RUN npm install -g llm-checker

COPY analyze.sh /usr/local/bin/
RUN chmod +x /usr/local/bin/analyze.sh

CMD ["analyze.sh"]

🤝 Contributing

Local development

git clone https://github.com/developer31f/llm-checker.git
cd llm-checker
npm install
npm test
npm run dev check
npm link

Adding new models

Edit src/models/expanded_database.js:

{
  name: "New Model 7B",
  size: "7B",
  type: "local",
  category: "medium",
  requirements: { ram: 8, vram: 4, cpu_cores: 4, storage: 7 },
  frameworks: ["ollama", "llama.cpp"],
  quantization: ["Q4_0", "Q4_K_M", "Q5_0"],
  performance: {
    speed: "medium",
    quality: "very_good",
    context_length: 8192,
    tokens_per_second_estimate: "15-30"
  },
  installation: {
    ollama: "ollama pull new-model:7b",
    description: "Description of the new model"
  },
  specialization: "general",
  languages: ["en"],
  year: 2024
}

Improving hardware detection

Edit src/hardware/detector.js:

detectNewGPU(gpu) {
  const model = gpu.model.toLowerCase();
  if (model.includes('new-gpu')) {
    return {
      dedicated: true,
      performance: 'high',
      optimizations: ['new-api']
    };
  }
  return null;
}

Contribution guide

1. Fork the repository
2. Create a feature branch git checkout -b feature/awesome-feature
3. Commit your changes git commit -am 'Add awesome feature'
4. Push the branch git push origin feature/awesome-feature
5. Open a Pull Request

Reporting issues

Include the following:

llm-checker check --detailed --export json > debug-info.json
DEBUG=1 llm-checker check > debug.log 2>&1
uname -a
node --version
npm --version

📊 Roadmap

v2.1 (Q2 2025)

• 🔌 Plugin system for extensions
• 📱 Mobile optimisations
• 🌐 Optional web UI
• 📈 Usage metrics & analytics
• 🔄 Automatic model DB updates

v2.2 (Q3 2025)

• 🤖 More back‑end frameworks (MLX, TensorRT)
• ☁️ Local vs cloud auto comparison
• 🎯 Fine‑tuning advisor
• 📊 Historical performance dashboard
• 🔒 Enterprise mode

v3.0 (Q4 2025)

• 🧠 AI performance predictor
• 🔄 Multi‑model orchestration
• 📱 Mobile companion app
• 🌍 Advanced multilingual models
• 🚀 Public cloud integrations

🏆 Credits

Technologies used

• systeminformation – cross‑platform HW detection
• Ollama – local LLM management
• Commander.js – CLI framework
• Chalk – terminal colours
• Ora – elegant spinners

Communities & projects

• llama.cpp – efficient LLM inference
• Hugging Face – model & dataset hub
• Meta Llama – open‑source Llama models
• Mistral AI – European LLMs
• Google Gemma – open Gemma family

Contributors

• Pavel Chmirenko – Lead developer & maintainer

Special thanks

• The Ollama team for an amazing local LLM tool
• Georgi Gerganov for llama.cpp
• The open‑source community for making AI accessible

📄 License

MIT – see LICENSE for details.

💝 Support the project

If LLM Checker saves you time:

⭐ Star the repo
🐛 Report bugs & suggest features
🤝 Contribute code or docs
📢 Share with fellow devs
☕ Buy me a coffee