Chinese artificial intelligence startup DeepSeek released a preview version of its new model, the V4, on Friday. The long-awaited next-generation model features an ultra-long context of one million words.
The DeepSeek-V4 is available in a pro version and a cheaper flash version. V4-Pro has 1.6 trillion parameters while the V4-Flash has 284 billion parameters, which determine models' decision-making ability.
In world knowledge benchmarks, DeepSeek-V4-Pro significantly leads other open-source models and is only slightly outperformed by the top-tier closed-source model, Gemini-Pro-3.1.
Releasing the preview versions allows the company to incorporate real-world feedback before finalizing the model.