Apex Intelligence Releases 35B Model Apex-35B-V1.0, Outperforming Larger Rivals

Apex Intelligence, founded by Chen Yongchao, a 27-year-old assistant professor at Tsinghua University's School of Artificial Intelligence, released its first 35B-parameter model, Apex-35B-V1.0, on October 28, 2026. The model is designed to balance general-purpose and research capabilities, handling tasks such as coding, agent interactions, Q&A, and deep search, while also supporting AI research, computational science, and robot simulation.

In 21 benchmark tests, Apex-35B-V1.0 led the same-scale Qwen3.6-35B-A3B in 20 categories. Against Recursive's Flint (40B), it achieved 18 wins and one tie. In some challenging evaluations, the model surpassed much larger models with 500B and 800B parameters.

According to the technical blog, Apex-35B-V1.0 was built through continued training and post-training on Qwen3.6-35B-A3B, incorporating proprietary datasets and research traces. The company integrated recursive self-improvement (RSI) into model development, allowing AI to participate in experiments and result analysis, and applying validated improvements to subsequent R&D. Its automated research system, Apex Research, covers training data construction, automated experimentation, training strategy exploration, and GPU kernel and inference optimization.

Founded in June 2026, Apex Intelligence completed two funding rounds totaling nearly 400 million yuan within four months before releasing its first competitive model.