Tencent backs Naive AI, valuing the Chinese startup at $1.42 billion after just seven months

A company called Naive AI, founded barely seven months ago, is now worth $1.42 billion. Tencent helped make that happen, joining a roster of investors betting that the future of AI isn’t about building bigger models. It’s about making the ones that already exist actually work better.

The Beijing-based startup has raised roughly $400 million across three funding rounds since its February 2026 launch.

The anti-foundation model play

Instead of pretraining new models, the startup focuses on improving existing open-weight AI models. It uses techniques like mid-training, post-training, and reinforcement learning to squeeze more performance out of architectures that are already publicly available.

The company plans to release its first open-weight large language model as early as September 2026. That timeline, if it holds, would mean the startup went from founding to shipping a model in about seven months, all while operating with fewer than 100 employees.

The investor lineup reads like a who’s who of Chinese tech capital. Alongside Tencent, the rounds included IDG Capital, MPCi, and HSG, the firm formerly known as Sequoia Capital China.

From $800 million to $1.42 billion in months

The funding came in three distinct tranches: $100 million, then $180 million, then $120 million. Back in April 2026, after roughly $300 million had been raised, the company’s valuation sat around $800 million. The latest round nearly doubled that figure to $1.42 billion.

Naive AI isn’t alone in this trend. Another lab, led by Lin Junyang, a former leader of Alibaba’s Qwen team, recently closed its first funding round at a valuation of approximately $2 billion post-money.

Founder Dai Jifeng brings serious credentials to the table. He’s an associate professor at Tsinghua University and previously held roles at Microsoft Research Asia and SenseTime, one of the country’s largest computer vision companies.

His background isn’t without controversy, though. Dai’s departure from MiroMind in early 2026 involved public disputes over intellectual property and operational disagreements.

Why optimization over pretraining matters

Open-weight models, released by companies like Meta with its Llama series and Alibaba with its Qwen models, have created a vast library of capable base models that anyone can access.

Tencent’s involvement is particularly telling. The Chinese tech giant has been consistently backing AI startups across the country, and its participation in Naive AI signals a strategic interest in companies pursuing scalable, efficient approaches to AI development. The company operates one of China’s largest cloud platforms and messaging ecosystems through WeChat, both of which stand to benefit from better AI models regardless of who trains them.