Event

Chaoyan Intelligence releases and open-sources 35B model Apex-35B-V1.0

2 reports · 2 sources

Chaoyan Intelligence, founded by Chen Yongchao, Assistant Professor at Tsinghua University's School of Artificial Intelligence, has released its first 35B model, Apex-35B-V1.0, and opened the model weights under the Apache 2.0 license; the official release provides approximately 72GB of BF16 weights and GGUF quantized versions at different precisions. The model has a total parameter count of 35 billion, activates approximately 3 billion parameters per token, supports a 262,000-token context, and is based on Qwen3.6-35B-A3B with optimization through mid-training and post-training, incorporating Chaoyan Intelligence's own datasets and research trajectories, balancing general capabilities such as Coding and Agent with AI research, computational science, and robot simulation tasks. According to Chaoyan Intelligence, across 21 benchmark tests, Apex-35B-V1.0 leads the comparable-scale Qwen3.6-35B-A3B on 20; compared with the 40B Recursive Flint, it leads on 18 and ties on 1, reaching SOTA level. In the RoboRSI evaluation, the model scored 44.5, versus 35.0 for Qwen3.6-35B and 30.0 for Flint. The above results are all claims by the releasing party.