Chinese AI maker DeepSeek has unveiled its new experimental model, DeepSeek-V3.2-Exp, claiming it can slash API costs by more than 50% compared to market rivals like OpenAI. The move is expected to intensify competition in the global AI race.
DeepSeek, based in Hangzhou, has already shaken Silicon Valley with its earlier V3 and R1 models, which gained attention for delivering high performance at low cost. The company’s strategy centers on making AI capabilities more affordable while pushing technological efficiency.
According to a post on Hugging Face, DeepSeek described its new model as an “intermediate step toward our next-generation architecture.” The upcoming architecture is expected to be the firm’s biggest launch yet.
The DeepSeek-V3.2-Exp comes with Sparse Attention technology, aimed at reducing computing costs and improving performance on long text sequences.
Analysts believe while this launch may not create the same market shock as January’s breakthroughs, it could still pressure rivals like Alibaba’s Qwen and OpenAI by offering competitive performance at far lower prices.
Experts say DeepSeek’s formula of “high capability + low cost” continues to make it one of the most closely watched AI developers worldwide.