Connecting Open-Weight Model Gateways
Enterprises often want to use open-weight models — like Llama or Mistral — instead of, or alongside, proprietary models from providers like OpenAI and Anthropic. These models are typically accessed through a gateway provider rather than directly from the model's creator, and until now, connecting one of these gateways to elvex wasn't straightforward.
elvex supports a set of gateway providers that give you access to open-weight and multi-model catalogs through a single connection.
You might need this when:
Your organization wants to use open-weight models for compliance, cost, or flexibility reasons
You want to access a wide catalog of models — including open-weight ones — through a single provider instead of connecting each model creator individually
You want to compare performance, latency, or cost across different hosting options using the same agent configurations
Supported Gateway Providers
elvex supports the following gateway providers, which give you access to open-weight and multi-provider model catalogs:
Meta
OpenRouter
Vercel AI Gateway
Fireworks AI
HuggingFace
Cloudflare AI Gateway
These are in addition to elvex's first-party model providers (OpenAI, Anthropic, Google AI, Cohere, Mistral, Bedrock, xAI, Azure OpenAI, and Azure AI Foundry).
Benefits and Outcomes
Connecting a gateway provider allows you to:
Expand model choice — access open-weight models and multi-provider catalogs without connecting each model creator individually
Simplify integration — set up a new gateway provider in minutes using the same configuration interface as any other AI provider
Compare options — run the same agent against different gateways to evaluate performance, latency, and cost
Before You Begin
Only elvex Admin users have permissions to add AI providers
You'll need a valid API key from the gateway provider you want to connect
How to Connect a Gateway Provider
Navigate to Settings and select AI Providers.
Click Add a Provider.
Select the gateway provider you want to connect — for example, OpenRouter, Fireworks AI, HuggingFace, Vercel AI Gateway, Cloudflare AI Gateway, or Meta.
Enter a Provider Name (optional — defaults to the provider's name) and your API Key.
Save the configuration. Models available through that gateway will now appear in the model picker when configuring your agents.
Common Questions
Which gateway should I use?
It depends on which models and hosting terms fit your organization. Different gateways offer different model catalogs, pricing, and infrastructure — try connecting more than one to compare.
Do I need to configure each model individually?
No. Once a gateway provider is connected, the models it makes available appear directly in the model picker for your agents.
Can I use gateway providers alongside first-party providers like OpenAI or Anthropic?
Yes. You can connect as many providers as you need, and choose the right model — from any connected provider — for each agent.
What's Next
Once a gateway provider is connected, select any of its models from the provider dropdown when configuring an agent. See Adding an AI Provider for general setup details that apply to all providers.