Oh My Pi’s model catalog is one of its most impressive features — it supports virtually every major LLM provider. With v16.3.15, it just added Grok 4.5, xAI’s latest model.
What’s New
Two commits in v16.3.15:
feat(catalog): added Grok 4.5 model support— adds the model to the catalog with correct pricing, context limits, and capabilitiesfeat: added prompt-cache affinity support for grok models— enables prompt caching for Grok models, reducing token costs for repeated prompts
Why Prompt-Cache Affinity Matters
Prompt caching is a cost optimization technique. Instead of sending the same system prompt with every request, the provider caches it and charges less for repeated prefixes. Without prompt-cache affinity, the agent sends the full context every time. With it, only the new tokens are billed at full price.
For coding agents that use long system prompts (most of them), prompt-cache affinity can reduce costs by 30-50%.
The Broader Trend
Oh My Pi’s model catalog now includes:
- OpenAI (GPT-4, GPT-4o, o1, o3)
- Anthropic (Claude 3.5, Claude 4)
- Google (Gemini 2.5)
- xAI (Grok 2, Grok 3, Grok 4, Grok 4.5)
- Open-source (Llama, Mistral, Qwen via Ollama/compatible)
- Baseten (hosted open models)
The catalog is one of the largest among coding agents. Each model is tested for compatibility, pricing is tracked, and routing is automatic.
How to Use Grok 4.5
In Oh My Pi, select Grok 4.5 from the model picker during setup or use:
/mode grok-4.5The model is available through xAI’s API with your existing API key.