DeepSeek Launches V4.1 Flash, Intensifying AI API Pricing Wars

DeepSeek has released V4.1 Flash, a 552 billion parameter open-weight model that reportedly outperforms the company's previous flagship, V4 Pro, on most benchmarks at a significantly lower cost. This release, along with Sakana's introduction of tiered pricing for its Fugu model, has led to an approximate order-of-magnitude reduction in the cost of frontier-adjacent coding and agent capabilities within 48 hours. DeepSeek plans to retire V4 Pro in two days, with V4.1 Flash serving all requests at its new, lower prices.