Chinese AI startup DeepSeek, which shook the tech world last year with its low-cost chatbot, released a preview version of its highly anticipated V4 model on Friday, December 20.

The company announced on social media that the new model "features an ultra-long context of one million words," enabling it to process vast amounts of text simultaneously.

DeepSeek-V4 comes in two versions: V4-Pro with 1.6 trillion parameters and the cheaper V4-Flash with 284 billion parameters. Parameters determine a model's decision-making ability.

According to the company, "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, Google's Gemini-Pro-3.1."

Releasing preview versions allows the company to incorporate real-world feedback before finalizing the model.

DeepSeek caused a stir in January 2025 when it unveiled a generative AI chatbot rivaling US products like ChatGPT but claiming significantly lower computing power and development costs.

However, the startup has also attracted controversy. Its chatbot reportedly dodges questions on politically sensitive topics such as the 1989 Tiananmen crackdown, raising censorship concerns.

The Hangzhou-based startup has also been accused by the United States and its American competitors of improper and illegal conduct. On Thursday, the White House alleged Chinese entities were engaging in "industrial-scale distillation campaigns to steal American AI."

Beijing rejected what it called "baseless allegations," adding that China "attaches great importance to the protection of intellectual property rights."

Source: www.dw.com