DeepSeek V4 ships 1M context, MIT-licensed open weights — but markets shrug

DeepSeek shipped V4-Pro and V4-Flash preview models on April 24 under MIT-licensed open weights with 1M context, undercutting Western API prices by an order of magnitude. Markets responded coolly — Omdia's Lian Jye Su called the announcement predictable as architecture and efficiency gains have become widely replicated.