Google Plans Massive $40 Billion Investment in Anthropic
April 26, 2026
DeepSeek launches V4 AI model with ultra-long context and lower costs, intensifying global AI competition and expanding commercial accessibility.
Chinese startup DeepSeek has unveiled its latest artificial intelligence model, DeepSeek-V4, marking a significant advancement in the global AI race. The new model introduces an ultra-long context window of one million words while significantly reducing compute and memory costs, positioning it as a competitive alternative to leading Western AI systems.
The release follows DeepSeek’s earlier breakthrough with its R1 reasoning model, which challenged assumptions about US dominance in artificial intelligence. With V4, the company aims to further enhance performance across reasoning, world knowledge, and agent-based capabilities.
DeepSeek-V4 is available in two versions: V4-Pro, featuring 1.6 trillion parameters, and V4-Flash, a more efficient and cost-effective variant. The model has also been optimised for integration with popular AI agent tools, supporting broader adoption across industries.
Experts describe the launch as a potential turning point, highlighting improvements in handling long-context tasks while lowering operational costs. The development could enable wider commercial use of advanced AI beyond research environments.
The announcement comes amid rising geopolitical tensions over artificial intelligence, with US officials alleging large-scale technology acquisition efforts by Chinese entities, further intensifying global competition in the sector.
