LLM Configuration

Select and configure the language model for the phone assistant

Active Model

Claude Sonnet 4

Anthropic

Available Models

Claude Opus 4

Anthropic

Most capable model. Best for complex reasoning, nuanced conversations, and high-stakes interactions.

mediumhigh200K tokens

Complex reasoningNuanced toneLong context

Claude Sonnet 4

Recommended

Anthropic

Balanced performance and speed. Ideal for real-time phone conversations with good quality.

fastmedium200K tokens

Fast responsesGood qualityCost-effective

Claude Haiku 4

Anthropic

Fastest model. Best for high-volume, low-latency phone calls where speed matters most.

fastlow200K tokens

Ultra-fastCheapestHigh throughput

Qwen 2.5 72B

OpenRouter (Free)

Free open-source model via OpenRouter. Good multilingual support including German.

mediumfree128K tokens

Free tierMultilingualOpen source

Gemini 2.0 Flash

OpenRouter (Free)

Google's fast model via OpenRouter free tier. Good for straightforward conversations.

fastfree1M tokens

Free tierVery fastLarge context

Note: Model selection and system prompt changes are currently display-only. To change the active model, update the ASSISTANT_MODEL environment variable in the assistant service configuration and restart.

A future update will enable live model switching and prompt persistence from this page.