DeepSeek has significantly increased API prices for its V4 Pro model, particularly for "cache hit" inputs, with some rates rising up to 11 times. This price hike reflects the growing costs associated with storing and moving data for large language models, which are becoming more expensive than computation itself. The increased demand for long context windows has led to "cache thrashing" on DeepSeek's servers during peak times, where data is frequently moved between GPU memory and slower storage, negating the benefits of caching and driving up operational expenses. While DeepSeek remains a cost-effective option compared to competitors like Anthropic, the price adjustment signals a shift towards developers needing to manage caching strategies more carefully to optimize costs. AI
IMPACT Signals a shift in AI infrastructure costs, requiring developers to manage caching strategies more actively.
RANK_REASON Price increase for a major AI model API due to infrastructure cost shifts. [lever_c_demoted from significant: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
