OpenClaw (or similar AI coding/agent frameworks) now support many model providers through a pluggable architecture. The list you see is basically LLM providers / gateways you can configure as the backend for the agent.
I’ll explain them in simple categories so you understand which ones are actually useful.
1️⃣ Major Frontier Model Providers (Most Powerful)
These provide the best reasoning and coding models.
| Provider | Example Models | Use Case |
|---|---|---|
| OpenAI | GPT-4.1, GPT-5 series | Coding, reasoning, agents |
| Anthropic | Claude Opus / Sonnet | Long context, coding |
| Gemini 1.5 / Gemini 2 | multimodal + long context | |
| xAI | Grok | real-time knowledge |
| Mistral AI | Mixtral / Mistral Large | EU based models |
These are usually best quality models.
2️⃣ Chinese Frontier Models
Very powerful and often cheaper.
| Provider | Models |
|---|---|
| Moonshot AI | Kimi K2.5 |
| Alibaba Cloud | Qwen 2.5 |
| Baidu | Qianfan |
| BytePlus | Doubao models |
| Volcano Engine | LLM services |
| MiniMax | MiniMax models |
| Zhipu AI | Z.AI / GLM |
These are popular for self-host or cheap inference.
3️⃣ Aggregators (Multi-Model Gateways)
These platforms let you access many models with one API.
| Provider | What it does |
|---|---|
| OpenRouter | access 100+ models |
| Together AI | open-source model hosting |
| Hugging Face | thousands of models |
| LiteLLM | unified API for all models |
| Cloudflare | LLM routing + caching |
| Vercel | AI routing for apps |
These are useful when building multi-model agents.
4️⃣ Local / Self-Hosted Models
For on-prem inference.
| Provider | What it is |
|---|---|
| vLLM | High-performance inference engine |
| Synthetic | private LLM infrastructure |
| Custom Provider | your own API |
Example:
OpenClaw
→ LiteLLM
→ vLLM
→ Llama3
5️⃣ Developer Platform Providers
These integrate AI with coding tools.
| Provider | Purpose |
|---|---|
| **Microsoft Copilot | coding models |
| Kilo Gateway | AI gateway |
| OpenCode Zen | coding LLM |
6️⃣ What Most Engineers Actually Use
In real production systems the common stack is:
Primary model
→ OpenAI
→ Anthropic
→ Google Gemini
Fallback models
→ Mistral
→ Grok
→ Qwen
Gateway
→ LiteLLM
→ OpenRouter
7️⃣ Best Setup for an AI Coding Agent
For something like OpenClaw / Codex-style agents I recommend:
Primary model:
OpenAI GPT-4.1 / GPT-5
Coding fallback:
Claude Sonnet
Cheap background tasks:
Mistral / Qwen
Gateway:
LiteLLM
💡 Based on your workflow (DevOps, infrastructure, Terraform, Kubernetes, coding automation), the best combination for you would be:
LiteLLM gateway
├── OpenAI
├── Anthropic
├── Google
└── Mistral
This gives best reasoning + redundancy.