DeepSeek V4 Flash Trains on 90 Trillion Tokens for $920K, Goes Offline Due to Heat
DeepSeek
DeepSeek V4 Flash, a new AI model, was trained on 90 trillion tokens at a cost of only $920,000. The model gained such immense popularity that its servers went offline due to overwhelming demand, according to a report by Phoenix New Media Technology.
DeepSeek V4 Flash, the latest AI model from DeepSeek, was trained on a staggering 90 trillion tokens for a mere $920,000. The remarkably low training cost is attributed to efficient architecture and optimized training techniques. Upon release, the model caused a frenzy among users, leading to server overloads and temporary outages. The incident underscores the growing demand for high-performance yet affordable AI models.
Source: DeepSeek (GNews) —
original