China's AI startup DeepSeek will unveil its next flagship AI model, "V4", within weeks, The Information reported on Jan. 9 local time, citing two sources familiar with internal matters.

DeepSeek's V4 is the successor to V3, released in December 2024, and will have strong coding capabilities, The Information reported.

Internal benchmark-based tests by DeepSeek employees showed V4 outperformed existing models such as Anthropic's Claude and OpenAI's GPT series in coding capability, the sources said.

DeepSeek plans to release V4 around the Lunar New Year in mid-February, but the launch schedule could change, The Information reported.

After releasing V3 in December 2024, DeepSeek introduced "R1", an open-source reasoning model focused on solving complex problems, in January 2025. If V3 marked the moment DeepSeek made its presence known in the global AI community, R1 sent shockwaves through Silicon Valley and Wall Street and had a large enough impact to trigger the DeepSeek phenomenon. DeepSeek also launched a chatbot in China using the R1 and V3 models.

DeepSeek released the V3.2 model in December last year. V3.2 outperformed OpenAI's GPT-5 and Google's Gemini 3.0 Pro on some benchmarks, The Information reported.

V4 is not a small upgrade based on V3 but a next-generation flagship model, and attention is on its competition with models from OpenAI, Google and Anthropic that lead the global market.

According to The Information, V4 made significant progress in handling and processing very long coding prompts. That could be a major advantage for developers working on complex software projects, The Information reported. V4 is also reported to have improved its ability to identify data patterns across every stage of training without performance degradation.