DeepSeek launches V4 Pro as China releases 10 AI models in 70 days

Nine major Chinese AI firms have released 10 new models and major updates over roughly 70 days since June 2026, marking an intensified phase of competition that centers on technical sophistication rather than cost alone. The burst of releases culminated on August 13 when DeepSeek formally launched its V4 Pro model, the latest in a rapid-fire sequence that has reshaped perceptions of Chinese AI capabilities.reuters

A Wave of Releases

The succession began with Zhipu AI's GLM-5.2 and was followed by Meituan's LongCat-2.0, Tencent's Hy3, Moonshot AI's Kimi K3, and Alibaba's ↘ 2.44% Qwen3.8-Max, according to Seoul Economic Daily. Moonshot AI's Kimi K3, unveiled in mid-July, stands out as the largest open-weight AI system released this year at 2.8 trillion parameters, and has ranked ahead of some leading U.S. models in composite evaluations.sedaily+2

On the Artificial Analysis Intelligence Index v4.1.1, DeepSeek V4 Pro scored 53, trailing GPT-5.6 Sol at 61 and Claude Opus 5 at 63 — but at an average cost per task of $0.06, one-39th the cost of Claude Opus 5's $2.34. The gap in raw capability is narrowing even as the cost advantage remains dramatic.sedaily

Global Companies Take Notice

The advances are reshaping how multinational corporations engage with Chinese technology. Apple ↗ 1.00% has tapped Alibaba and Baidu ↘ 0.15% for AI services in China, while Ford ↗ 0.43% is working with CATL Contemporary Amperex Technology Co., Limited on battery technology for a $3.5 billion plant in Michigan, CNBC reported.cnbc

An IDC survey of European companies earlier this year found that security, compliance, and superior performance — not cost — were the top reasons for adopting Chinese AI models. "The popular narrative that Western companies are rushing to Chinese AI because it's cheap gets this backwards," Kitty Fok, managing director at IDC China, told CNBC. "The decision is performance-led and compliance-gated."cnbc

Structural Shift, Strategic Tension

The rapid improvement of Chinese open-source models has prompted calls within the U.S. to bolster its own open-source ecosystem. Meta ↗ 2.78% CEO Mark Zuckerberg published a manifesto this week arguing that restrictions on U.S. AI companies are handing advantages to foreign competitors.global.chinadaily

Lian Jye Su, chief analyst at Omdia, told CNBC that U.S. export controls have paradoxically "become a catalyst for Chinese domestic innovation and efficiency." The result, analysts say, is a more fragmented but pragmatic global technology ecosystem where Chinese models lead in cost-effectiveness and increasingly compete on raw performance — a dynamic that neither sanctions nor corporate caution have so far reversed.cnbc