Why route across multiple LLMs
No single model is best at everything, and committing to one provider is a business risk. Pricing, availability and capability all change. Routing lets you match each task to a model and change your mind later without re-architecting.
The 10 model providers
- Anthropic
- OpenAI
- Google Gemini
- Perplexity
- Groq
- OpenRouter
- Together
- Qwen Cloud
- Ollama (local)
- vLLM (local)
Cloud and local models sit behind one interface. Ollama and llama.cpp cover execution on your own hardware. The rest cover hosted models.
Routing and governance together
The point that matters to a business is that routing does not bypass policy. In governed mode the governance check runs first, then cost preflight, then model routing. Model flexibility does not cost you oversight.
Questions
What is multi-LLM routing?
Multi-LLM routing sends each task to the model that fits its cost, capability and policy needs, instead of hardwiring one provider. Daena routes across 10 model providers.
Which models can Daena route across?
10 model providers: Anthropic, OpenAI, Google Gemini, Perplexity, Groq, OpenRouter, Together, Qwen Cloud, Ollama (local), vLLM (local). Local models run on your own hardware through Ollama or llama.cpp. Keeping a whole deployment local also depends on its connectors, logs and backups.
Can I swap one provider for another without breaking my agents?
That is the point of a routing layer: routing and policy live in the platform, not in each agent. The details of a migration depend on your prompts and tools, so test the swap on your own workloads before you rely on it.
Does switching models change governance?
No. In governed mode the governance check runs before model routing, so every request is checked whichever model ends up handling it.
How does Daena decide which model to use?
Through three routing modes. Standard picks the best single model. Council sends the task to three or more models and synthesizes one answer. Quintessence pairs experts with models in a matrix. A cost preflight stage runs before routing.