DeepSeek has launched grayscale testing for its new model on both its official website and app. 1M content length!
A massive 1M context window could change everything for long-form AI tasks...
Deep Dive
DeepSeek has launched grayscale testing for a powerful new model, reportedly outperforming Google's Gemini 2.5 Pro in early benchmarks without web search. The model features a massive 1 million token context window and an updated knowledge base. Access is currently limited to a select group of accounts via the official website and app, with the community speculating it may be a 'V4 Lite' version rather than the full V4 release.
Why It Matters
This directly challenges Google's flagship model and could significantly lower the cost of ultra-long-context AI applications.