Chinese AI firm DeepSeek is delaying the release of its R2 model, the successor to its successful R1 reasoning model, according to a report by The Information. CEO Liang Wenfeng is reportedly unsatisfied with R2’s current performance, leading to an indefinite postponement.
Originally slated for a May release, R2 was intended to improve upon R1’s capabilities, particularly in coding and multilingual reasoning beyond English. For months, DeepSeek’s engineers have been focused on refining the model in anticipation of Liang’s approval. The delay highlights the company’s commitment to quality and its ambition to deliver a superior AI offering. The exact timing for R2’s launch remains uncertain as DeepSeek continues to strive for optimal performance.
