DeepSeek Slashes AI Costs to Cents
5 hours ago
- #enterprise AI
- #AI pricing
- #DeepSeek V4 Pro
- DeepSeek made the 75% price reduction on its V4 Pro model permanent, lowering costs to $0.435/M input and $0.87/M output tokens.
- The V4 Pro model uses a Mixture-of-Experts architecture with 1.6 trillion parameters and an MIT license, enabling full commercial use.
- Its efficiency and caching features, like cache-hit input at $0.003625/M, allow for significant cost savings in enterprise AI workloads.
- Techniques like prefix caching and using V4-Flash as a router can further optimize token costs in production RAG pipelines.
- Other news includes AI models for disease prediction, China's automotive standardization plan, and Altman's view that AI won't cause a jobs apocalypse.