Choose an AI model for the work your app does
Build with Claude, GPT, Grok, and open models through an OpenAI-compatible endpoint. Compare model capabilities and pricing before trying them on your own prompts.
Match the model to your task
Start with what your app needs to do. A model that summarizes short messages may have different requirements from one that reads images or calls tools. Check the context window and supported features before comparing prices.
The model catalog returns those details alongside input and output token prices. Use it to build a shortlist, then test your prompts and compare both the answers and their cost. Where we have measured results, the leaderboards include test dates and methodology.
The shared inference endpoint uses the OpenAI chat-completions format. You select a model with the model field, so you can try another compatible model in the same client. Provider-native routes are also available to enabled accounts during the pilot for supported provider-specific features.
Explore the guides
Compare providers
Compare starting prices and the tasks each provider supports. Features and usage affect the total cost. For measured results, see the benchmarks and how we test.
| Provider | Best suited for | Starting price |
|---|---|---|
| Anthropic | Use Claude for text, image, and tool-based tasks. | metered by Anthropic |
| Grok | Use Grok models for supported inference tasks. | metered by Grok (xAI) |
| Hugging Face | Try supported open models for your workload. | metered by Hugging Face |
| Jina | Add embeddings, reranking, or text classification. | $0.0001 per call |
| OpenAI | Use GPT models for supported text, image, and tool tasks. | metered by OpenAI |
Use it with NativePort
Try models from different providers using one NativePort key. Your shared balance also covers the search and scraping APIs your agent uses.
Before you start
Can I keep my existing client library?
If it supports the OpenAI chat-completions format and a custom base URL, point it at /inference/v1 and use your NativePort key. The integration guides cover supported frameworks and their setup.
Can I use a provider's native API?
The /anthropic, /openai, and /grok routes are enabled per account during the pilot. An account without access receives a 404 on those routes. The shared inference endpoint is available to every account and is the starting point for these guides.
Where can I check model capabilities and prices?
GET /inference/v1/models lists available models with their context windows, capabilities, and USD token prices. The model-pricing guide below shows how to estimate the cost of a workload before running it.