Context Caching on Disk

Context Caching on Disk — 90% price cut for cache-hit tokens

Why it matters

Productized the prefix cache; an internal serving optimization surfaced in the API price list

Source — Primary sources first; secondary reports are labelled “Press”

Officialapi-docs.deepseek.comhttps://api-docs.deepseek.com/news/news0802/

Added in Week 35, 2026 · See that week

Same week

AI Act enters into forcePressThe Llama 3 Herd of Models publishedPaper

Nearby on the same axis

Retrieval Augmented Generation or Long-Context LLMs? publishedPaperAnthropic launches prompt caching in public betaOfficialColPali publishedPaperGoogle opens Gemini 1.5 Pro's 2 million-token context to all developersOfficial