Providers & Models
OpenAI
API base: https://api.openai.com/v1
Dashboard: platform.openai.com
| Model | Notes |
|---|---|
gpt-4o | Best quality, recommended for support |
gpt-4o-mini | Faster and cheaper, good for simple tickets |
gpt-4-turbo | Previous generation flagship |
gpt-3.5-turbo | Cheapest option, lower quality |
Anthropic (Claude)
API base: https://api.anthropic.com/v1
Dashboard: console.anthropic.com
| Model | Notes |
|---|---|
claude-opus-4-7 | Most capable, highest cost |
claude-sonnet-4-6 | Balanced — recommended for support |
claude-haiku-4-5-20251001 | Fastest and cheapest |
claude-3-5-sonnet-20241022 | Previous generation, widely used |
claude-3-5-haiku-20241022 | Previous generation fast model |
Mistral AI
API base: https://api.mistral.ai/v1
Dashboard: console.mistral.ai
| Model | Notes |
|---|---|
mistral-large-latest | Most capable Mistral model |
mistral-small-latest | Good quality at lower cost |
open-mistral-nemo | Open-weight model |
codestral-latest | Optimised for technical/code questions |
Custom (OpenAI-compatible)
Any API that speaks the OpenAI chat completions format can be used. Set the Custom API Base URL to your endpoint's base path.
Examples:
- OpenCode:
https://api.opencode.ai/v1 - Local model via Ollama:
http://localhost:11434/v1 - Together AI:
https://api.together.xyz/v1
The model name must match exactly what your endpoint expects.
note
When using a local or self-hosted model, make sure the Paymenter server can reach the endpoint. A model that is too slow (over 45 seconds to respond) will time out and trigger the escalation flow.