Models & routing

Rescriptum connects directly to OpenRouter, granting instant access to over 300 frontier and open-source models with zero provider markup.

Dynamic Model Switching

You can switch models at any turn in a conversation. Rescriptum automatically normalizes system messages and tool call formats so switching between Anthropic (Claude), OpenAI (GPT-4o), Google (Gemini), and DeepSeek is completely seamless.

Prompt Caching & Cost Optimization

Rescriptum optimizes prompts for provider-level prefix caching (e.g. Anthropic prompt caching). Static system instructions and codebase index summaries are pinned at the top of context, slashing cost by up to 80% on long conversations.

Arena Scoring & Model Intelligence

The Models panel includes empirical benchmark data collected from real coding tasks. See latency percentiles, tool-calling reliability scores, and token economics before picking a model for your task.