AI Tool Selection Comparison
The AI section of this site covers dozens of tools. This page provides a horizontal comparison to help you quickly choose based on your needs. Click the corresponding links for installation and detailed usage of each tool.
Pricing information may change with official policies. The table below is for reference only; please refer to each product's official website for the latest information. "Local" means it can run offline on your own machine, and data never leaves your device.
AI Editors / IDEs
Suitable for developers who want "native AI integration in the editor."
| Tool | Form | Local/Cloud | Billing Model | Suitable For |
|---|---|---|---|---|
| Cursor | Standalone editor (VS Code fork) | Cloud model | Free tier + subscription | Those wanting an out-of-the-box AI IDE |
| Windsurf | Standalone editor | Cloud model | Free tier + subscription | Those preferring agent-style workflows |
| Zed | High-performance native editor | Cloud model / can connect to local | Open source, AI features with free tier | Those seeking lightweight speed |
AI Coding Assistants (CLI)
Coding agents that work in the terminal, suitable for command-line lovers and cross-file modifications.
| Tool | Backend Model | Local/Cloud | Billing Model | Notes |
|---|---|---|---|---|
| Claude Code | Anthropic Claude | Cloud | Subscription or API billing | Terminal-native agent, excels at large changes |
| Gemini CLI | Google Gemini | Cloud | Free tier + API billing | Generous free quota |
| GitHub Copilot | Multiple models | Cloud | Subscription (personal/enterprise) | Mature ecosystem, wide IDE integration |
| Aider | Can connect to any model / local | Cloud or local | Open source (billed by model used) | Deep Git integration |
| Cline | Can connect to any model / local | Cloud or local | Open source (billed by model used) | VS Code extension form |
| Continue | Can connect to any model / local | Cloud or local | Open source (billed by model used) | Highly customizable |
| OpenCode | Can connect to any model / local | Cloud or local | Open source | Terminal coding agent |
Local Large Model Runtimes
Run large models offline on your own machine—data never leaves your device, suitable for privacy-sensitive users or those wanting to save API costs. Requires some memory/VRAM.
| Tool | Positioning | Ease of Use | Notes |
|---|---|---|---|
| Ollama | Pull and run models with one command | ⭐ Easiest | First choice for local models, wide ecosystem |
| LM Studio | Graphical interface | ⭐ Easy | Closed-source free, suitable for non-command-line users |
| Jan | Graphical interface, open source | ⭐ Easy | Open-source alternative to LM Studio |
| llama.cpp | Low-level inference engine | ⭐⭐⭐ Advanced | Ultimate performance, base for many tools |
| llamafile | Single executable model file | ⭐⭐ Medium | Packages the model into one file |
| LocalAI | Local service compatible with OpenAI API | ⭐⭐ Medium | Local backend for existing applications |
AI Image and Speech
| Tool | Purpose | Local/Cloud | Notes |
|---|---|---|---|
| Stable Diffusion WebUI | Text-to-image | Local | Most feature-rich, requires dedicated GPU |
| ComfyUI | Node-based text-to-image workflow | Local | Visual, orchestrate complex workflows |
| OpenAI Whisper | Speech-to-text | Local | Multilingual recognition, offline-capable |
AI Platforms and Automation
| Tool | Purpose | Deployment | Notes |
|---|---|---|---|
| Dify | LLM application development platform | Self-hosted | Build RAG / agent applications |
| n8n | Workflow automation | Self-hosted | Integrate AI into automation workflows |
| OpenClaw | Agent automation | Self-hosted | Task-based AI automation |
How to Choose? By Scenario
- Want to write code in an editor, most hassle-free → Cursor or Windsurf.
- Heavy command-line user, need cross-file major changes → Claude Code; want more free usage → Gemini CLI.
- Want open-source assistant with your own model / local model → Aider, Cline, or Continue.
- Privacy-conscious / want offline / save API costs → Use Ollama (command-line) or LM Studio/Jan (graphical) to run local models.
- Pursue extreme inference performance, need low-level customization → llama.cpp.
- Do text-to-image → Start with Stable Diffusion WebUI, use ComfyUI for complex workflows.
- Build AI applications / automation → Dify or n8n.
Hardware Tips
- Cloud tools (Cursor, Claude Code, Copilot, etc.) have low local hardware requirements but need internet, and data is sent to the service provider.
- Local models require memory/VRAM: A 7B-level model may work with about 8GB RAM; larger models are recommended with an NVIDIA dedicated GPU (see NVIDIA Optimus dual graphics configuration).
Further reading: AI Tools Overview · Ollama Local Large Models