DeepSeek announced on Wednesday (12) the creation of a new team focused on the market for AI coding agents and the update of the V4 Pro model, which now has better agent capabilities, PYMNTS reported.
The company posted job openings for the new team, which is expected to develop cutting-edge products for agents. It also created an official WeChat account for the "DeepSeek Harness Team", a name that refers to the software component that helps AI agents execute complex tasks.
Competition in the sector
The move comes amid the race for AI programming tools. Meta launched Muse Code, a terminal agent in beta, on August 5; OpenAI announced in June the expansion of Codex with the acquisition of Ona; and Google is negotiating with startup Mechanize the hiring of talent and a non-exclusive technology licensing agreement.
On Thursday (6), DeepSeek resumed its 2nd funding round, aiming to raise about US$ 8 billion at a valuation of approximately US$ 74 billion. The move comes after the launch of V4 Flash, a model that drew attention for its cost-effectiveness.
DeepSeek completed its first round in late May, with US$ 7 billion and a valuation of US$ 52 billion. The acceleration in its fundraising schedule is related to the expectation of more capital to build its own data center and acquire more chips.
DeepSeek gained prominence in January 2025 by launching a model with performance comparable to that of American rivals such as OpenAI and Meta, using substantially fewer Nvidia chips.
