Another Chinese MoE flagship lands third in the open-source ranks, with pricing aimed at DeepSeek-V4-Pro

On September 20, Zhidongxi reported that another Chinese MoE (mixture-of-experts) flagship model has arrived, taking third place among open-source models worldwide while pricing itself against DeepSeek-V4-Pro.

MoE has become the default architecture for Chinese flagship models, letting different expert subnetworks participate in computation on demand so that total parameter counts can be balanced against the cost of a single inference pass. A flagship normally means a vendor's main model for complex tasks, and a third-place open-source position puts this release in the same conversation as the leading open-weight models.

Pricing is the more interesting part of the report. Bench-marking the new model directly against DeepSeek-V4-Pro suggests the competition among Chinese open models is shifting away from leaderboard position and toward inference cost and commercial adoption: where capabilities are comparable, whoever drives the per-call price lower is more likely to be picked up by developers and enterprise buyers.

The report, however, is so far built on two data points — rank and price. Details such as how the model performs on real tasks, and which leaderboard and evaluation conditions sit behind the third-place claim, still need to be disclosed before the ranking can be treated as comparable.

For the industry, a dense run of Chinese open-weight flagship launches means buyers will compare three curves at once: capability, price and deployment overhead. A high ranking only settles whether the model is usable; price and stability decide whether it is affordable and safe to run long term.

What to watch next is how stable this model proves on real workloads, whether third parties can independently reproduce the results, and whether its pricing pulls rivals into another round of follow-the-leader cuts. For developers and enterprise procurement teams, long-term support and ecosystem fit matter as much as rank and price.