The Chinese DoorDash is back again. This time, the company released another big MoE with 1.6T parameters. While the model itself is not the most capable for its size beyond benchmarks, it was trained entirely on Ascend 910s, making it the first non-Huawei, non-toy model trained entirely on Chinese accelerators. Other Chinese chips are mostly used for inference (if at all).
Specs
Params1.6T (~48B active)
LicenseMIT
Tags
Capability · Artificial Analysis
AA Index
25.8
Months Behind Frontier
11.7 mo
grok-4
10 Jul 2025
Adoption · Hugging Face
RAM @ 30d
Relative Adoption Metric: 0.01×. Below benchmark. Measured at 30 days.
Hugging Face Downloads
1.5K
last 30d
5K
all time
HF Likes
265
Relative Adoption Metric contextualizes downloads against the model's size bucket.
Related Models
