# Cut inference latency and cost with built-in LLM response caching

> We’ve added native caching for model responses across sync, async, and streaming APIs. You can configure TTL and cache storage to accelerate repeated prompts and reduce token spend — without building…

- Published: 2025-10-29
- Author: Agno Team
- Category: Changelog
- Canonical: https://agno-com-nine.vercel.app/articles/cut-inference-latency-and-cost-with-built-in-llm-response-caching
- Markdown: https://agno-com-nine.vercel.app/articles/cut-inference-latency-and-cost-with-built-in-llm-response-caching.md

We’ve added native caching for model responses across sync, async, and streaming APIs. You can configure TTL and cache storage to accelerate repeated prompts and reduce token spend — without building your own cache layer. This is opt-in and works out of the box across the platform.

**Details**

- Works across all API modes (sync/async/streaming)
- Configurable TTL and cache directory for control and portability
- Reduces latency for repeated prompts in production and evaluation pipelines

**Who this is for:** Teams optimizing cost, responsiveness, and throughput for high-volume or repetitive LLM workloads.