Chinese Labs Release DeepSeek V4 and Open-Weights Coding Models at Lower Costs

The Surge

Four Chinese AI labs - Z.ai, MiniMax, Moonshot, and DeepSeek - have released open-weights coding models within a 12-day period. DeepSeek V4 hits Western frontier capability on agentic engineering at significantly lower inference costs, though still lagging on NIST cross-domain benchmarks.

DeepSeek V4

DeepSeek V4 is the latest iteration of the DeepSeek coding model series. It demonstrates competitive performance on coding benchmarks while operating at a fraction of the cost of comparable Western models. This cost advantage is a key differentiator in the increasingly competitive AI market.

The Chinese AI Push

The simultaneous release of multiple coding models from Chinese labs signals a coordinated push to compete with Western AI providers. China has been investing heavily in AI development, and these releases demonstrate the country growing capabilities in specialized AI domains.

Cost Comparison

Chinese models typically offer significantly lower inference costs compared to Western alternatives. This makes them attractive for businesses that need to process large volumes of coding tasks without breaking the budget.

Implications for Developers

The availability of high-quality, low-cost coding models from Chinese providers gives developers more options when building AI-powered development tools. Having access to multiple providers through a unified API platform enables developers to choose the best model for each coding task based on quality, cost, and language support.