Enterprises using multiple AI models are underestimating failure rates by 2.25x
Enterprises using multiple AI models underestimate failure rates by 2.25 times because they ignore the co-failure ceiling, where all models in a pool simultaneously fail on the same prompts. A study of 67 frontier models from 21 providers found that orchestration success depends not on model disagreement rates but on the percentage of prompts where every model gives wrong answers at once.
Read full story →Databricks Closes $5 Billion Round at $190 Billion Valuation
Databricks raised $5 billion in funding at a $190 billion valuation, representing a 42% increase in six months. The company has achieved a revenue run-rate exceeding $7 billion with year-over-year growth above 80%.
Anthropic’s $2trn IPO is priced below what AI stocks already fetch
Anthropic investors project the company could reach a $2 trillion valuation in an autumn initial public offering, exceeding SpaceX's $1.77 trillion IPO valuation from June. Anthropic has not confirmed any listing plans or valuation figures itself.
DeepSeek built a Claude Code rival, then quadrupled its prices
DeepSeek increased API prices for DeepSeek-V4-Pro from $0.87 to $3.96 per million output tokens at peak rates starting August 16, while V4-Flash prices rose from $0.28 to $1.32. Off-peak rates were set at half the peak price for both models.