DeepSeek previews new AI model that ‘closes the gap’ with frontier models

Chinese AI lab DeepSeek has launched two preview versions of its latest major language model, DeepSeek V4a long-awaited update to last year’s V3.2 model and its accompanying R1 reasoning model that took the AI ​​world by storm.

The company says that both DeepSeek V4 Flash and V4 Pro are a combination of experts with context windows of 1 million tokens each – enough to enable the use of large codebases or documents in prompts. In the mix-of-experts approach, only a certain number of parameters are activated per task to reduce inference costs.

The Pro model has a total of 1.6 trillion parameters (49 billion active), making it the largest open-weight model available, bettering Moonshot AI’s Kimi K 2.6 (1.1 trillion), MiniMax’s M1 (456 billion), and more than double DeepSeek V3.2 (671 billion). The smaller V4 Flash has 284 billion parameters (13 billion active).

DeepSeek says that both models are more efficient and performant than DeepSeek V3.2 thanks to architectural improvements, and that they almost “close” the gap with current leading models, both open and closed, in terms of reasoning benchmarks.

The company claims its new V4-Pro-Max model outperforms its open-source peers in reasoning benchmarks, outperforming OpenAI’s GPT-5.2 and Gemini 3.0 Pro on some tasks. When coding competitive benchmarks, DeepSeek said the performance of both V4 models is “comparable to GPT-5.4.”

However, the models appear to lag somewhat behind frontier models in knowledge tests, particularly OpenAI’s GPT-5.4 and Google’s latest Gemini 3.1 Pro. This delay suggests a “development trajectory that trails state-of-the-art boundary models by approximately three to six months,” the lab wrote.

Both V4 Flash and V4 Pro only support text, unlike many of its closed-source peers, which support understanding and generating audio, video, and images.

Notably, DeepSeek V4 is much more affordable than any frontier model currently available. The smaller V4 Flash model costs $0.14 per million input tokens and $0.28 per million output tokens, undercutting GPT-5.4 Nano, Gemini 3.1 Flash, GPT-5.4 Mini, and Claude Haiku 4.5. The larger V4 Pro model, meanwhile, costs $0.145 per million input tokens and $3.48 per million output tokens, also undercutting Gemini 3.1 Pro, GPT-5.5, Claude Opus 4.7, and GPT-5.4.

The launch comes a day after the US accused China is attempting to steal the IP of US AI labs on an industrial scale using thousands of proxy accounts. DeepSeek itself has been accused by Anthropic and OpenAI of “distilling,” essentially copying, their AI models.