Provider tiers
Tier 1 — Direct API
| Provider | Models | Use case |
|---|---|---|
| OpenAI | GPT-5-nano, GPT-4o | General chat, code generation |
| Anthropic | Claude Sonnet, Claude Haiku | Complex reasoning, analysis |
| Google AI | Gemini Pro, Gemini Flash | Multimodal, fast responses |
Tier 2 — Cloud platforms
| Provider | Models | Use case |
|---|---|---|
| Azure AI Foundry | Hosted model endpoints | Enterprise deployments |
Tier 3 — Routers and local
| Provider | Models | Use case |
|---|---|---|
| OpenRouter | 200+ models | Access to any model |
| Ollama | Self-hosted LLMs | Full privacy, no API costs |
Model routing logic
kombify AI selects models based on:- Task complexity — Simple questions use fast, cheap models; complex tasks use more capable models
- User preference — You can pin a specific model in settings
- Cost budget — Monthly budget limits are respected
- Provider availability — Automatic fallback if a provider is down
Default model selection
| Task type | Default model | Reasoning |
|---|---|---|
| Simple chat | GPT-5-nano | Fast, very low cost |
| Code generation | Claude Sonnet | Strong code capabilities |
| Complex reasoning | GPT-4o / Claude Sonnet | Best overall performance |
| Quick summaries | Gemini Flash | Fast, cost-effective |
Configuration
Override the default model in AI Settings > Models or per-conversation in the model picker dropdown.Further reading
BYOK setup
Configure your own API keys
Configuration reference
All AI configuration options