AI Models
OpenClaw supports multiple AI providers. Choose the one that fits your needs and budget.
Anthropic
The most capable model for complex tasks. Excellent at coding, analysis, and nuanced conversations.
Input
$3.00/1M
Output
$15.00/1M
Context
200K tokens
Pros
- ✓ Best overall quality
- ✓ Excellent coding ability
- ✓ Strong reasoning
- ✓ Good at following instructions
Cons
- ✗ Higher cost
- ✗ Slower than smaller models
Anthropic
Fast and affordable option for simpler tasks. Great for quick responses and high-volume usage.
Input
$0.25/1M
Output
$1.25/1M
Context
200K tokens
Pros
- ✓ Very fast
- ✓ Low cost
- ✓ Good quality for price
- ✓ Same context window
Cons
- ✗ Less capable on complex tasks
- ✗ May miss nuances
OpenAI
OpenAI's flagship multimodal model. Good all-around performance with image understanding.
Input
$2.50/1M
Output
$10.00/1M
Context
128K tokens
Pros
- ✓ Multimodal (images)
- ✓ Fast
- ✓ Good all-around
- ✓ Large ecosystem
Cons
- ✗ Smaller context than Claude
- ✗ Variable quality
OpenAI
Budget-friendly option from OpenAI. Decent performance at a fraction of the cost.
Input
$0.15/1M
Output
$0.60/1M
Context
128K tokens
Pros
- ✓ Very affordable
- ✓ Decent quality
- ✓ Fast
Cons
- ✗ Limited capabilities
- ✗ May struggle with complex tasks
Google's fast model with massive context window. Free tier available for low usage.
Input
Free tier
Output
Then $0.075/1M
Context
1M tokens
Pros
- ✓ Huge context window
- ✓ Free tier
- ✓ Fast
- ✓ Multimodal
Cons
- ✗ Quality varies
- ✗ Less tested with OpenClaw
Amazon Bedrock
Access Claude, Llama, Mistral through AWS. Enterprise-grade security with existing AWS infrastructure.
Input
Varies
Output
by model
Context
Up to 200K
Pros
- ✓ AWS integration
- ✓ Multiple models
- ✓ Enterprise compliance
- ✓ Data privacy
Cons
- ✗ Requires AWS account
- ✗ More complex setup
OpenRouter
Unified API for all major providers. Access Claude, GPT-4, Llama, and more with one key.
Input
Pay-per-use
Output
No minimums
Context
Varies
Pros
- ✓ One API for all
- ✓ Easy switching
- ✓ Auto failover
- ✓ Price comparison
Cons
- ✗ Slight latency overhead
- ✗ Third-party dependency
Vercel AI Gateway
Run AI at the edge with automatic failover and caching. Great for Vercel deployments.
Input
Pass-through
Output
+ caching savings
Context
Varies
Pros
- ✓ Edge-optimized
- ✓ Auto failover
- ✓ Built-in caching
- ✓ Unified SDK
Cons
- ✗ Requires Vercel
- ✗ Limited to supported providers
Local (Ollama, LM Studio)
Run AI completely locally with no API costs. Supports Ollama, LM Studio, and more with model failover.
Input
Free
Output
Forever
Context
8K-128K tokens
Pros
- ✓ 100% free
- ✓ Complete privacy
- ✓ No internet needed
- ✓ Model failover
Cons
- ✗ Requires good hardware
- ✗ Lower quality than cloud models
- ✗ Setup complexity
Our Recommendation
For most users, we recommend starting with Claude 3.5 Sonnet for the best quality, or Ollama with Llama 3.1 if you want to run completely free and private.
You can switch models anytime in your OpenClaw configuration. Many users configure multiple models and switch based on the task.
Ready to Get Started?
Install OpenClaw and configure your preferred AI model.