DeepSeek V4 ships at 1.6T params with day-0 Blackwell support and aggressive price cuts

DeepSeek released V4 in preview on April 24 with two variants: V4-Pro at 1.6T parameters and V4-Flash at 284B, both supporting up to 1M-token context and using just 27% of single-token inference FLOPs and 10% of the KV cache versus prior frontier models. NVIDIA delivered day-0 Blackwell support hitting 3,500 tokens/sec on the 1.6T model via NVFP4, while DeepSeek slashed flagship pricing to pressure US rivals — though Reuters notes markets are no longer wowed.