Chinese AI models surge as DeepSeek prepares V4 launch amid open-source dominance

DeepSeek Sparked a Global Shift in AI Economics

When DeepSeek released its R1 reasoning model in January 2025, the impact extended far beyond technical circles. Within days, the Chinese AI startup's assistant app replaced ChatGPT as the most downloaded free app in the US App Store1

1

. The moment triggered a sharp sell-off that briefly erased roughly $1 trillion in market value across US tech stocks1

1

. Nvidia alone saw shares plunge 17%, wiping out $593 billion in a single day before recovering2

2

. The shock stemmed from DeepSeek's claim that it had built a model comparable to OpenAI's best at a fraction of the cost, challenging the assumption that only companies spending tens of billions of dollars on computing infrastructure could produce cutting-edge AI2

2

.

Chinese Open-Source AI Models Now Dominate Downloads

The competitive landscape has shifted dramatically. On Hugging Face, Alibaba's Qwen family has overtaken Meta's Llama models in cumulative downloads after ranking as the most downloaded model series in 2025 and 20261

1

. A recent MIT study found that Chinese open-source AI models have surpassed US models in total downloads1

1

. Moonshot AI's latest open-weight model, Kimi K2.5, came close to top proprietary systems such as Anthropic's Claude Opus on some early benchmarks while costing roughly one-seventh the price1

1

. These Chinese AI models differ crucially from most US systems by publishing their weights, allowing anyone to download, run, study, and modify them1

1

.

A Flurry of Low-Cost Chinese AI Models Expected Soon

One year after DeepSeek's breakthrough, intensifying competition among Chinese AI companies is producing a new wave of releases timed around China's Lunar New Year holiday, which officially begins on February 152

2

. Zhipu AI released its latest AI model on Wednesday, featuring enhanced coding capabilities and the ability to perform long-running tasks without user prompts2

2

. ByteDance officially unveiled Seedance 2.0 on Thursday, a video generation AI model capable of producing cinematic content in seconds, according to the Global Times2

2

. ByteDance is also expected to upgrade its Doubao chatbot, currently China's most popular AI app with 155.2 million weekly active users2

2

.

DeepSeek V4 Aims to Reclaim Technical Leadership

DeepSeek is preparing to release its next-generation model V4, which could replace the V3 model that powered the assistant app3

3

. The company recently upgraded its chatbot's context window, expanding how much information it can process and remember in a single task3

3

. V4 introduces four major technical upgrades: MODEL1 architecture with tiered KV cache storage aims to cut memory use by 40%, sparse FP8 decoding achieves a 1.8x inference speedup, Engram memory modules separate short-term context from long-term memory, and mHC optimized residual connections improve training efficiency by 30%3

3

. This release matters because DeepSeek's market share fell from 50% at the start of 2025 to under 25% by year-end as competition from Qwen, Kimi K2, and InternLM intensified3

3

.

Low Deployment Costs Reshape the Global Landscape of AI Innovation

A report by research group RAND on US-China AI competition found that Chinese models operate at roughly one-sixth to one-fourth the cost of comparable US systems2

2

. "DeepSeek showed the industry that you can create a very good model even when you're resource-constrained," said Lian Jye Su, chief analyst at tech research firm Omdia2

2

. "The combination of open-source access, strong reasoning capabilities and low deployment costs has become a defining model for how Chinese vendors now approach foundation models"2

2

. Before DeepSeek's breakout, some Chinese industry leaders, including Baidu CEO Robin Li, had argued that closed-source systems would dominate. Within days of DeepSeek overtaking ChatGPT in Apple's App Store downloads, Baidu and other leading firms began opening portions of their own models4

4

.

Democratizing Access to Advanced AI Through a Model + Tools Ecosystem

Hugging Face is now dominated by releases from Chinese tech giants such as Baidu, ByteDance and Tencent, and startups such as Moonshot4

4

. For developers and builders worldwide, access to near-frontier AI capabilities has never been this broad or this affordable1

1

. DeepSeek is shifting strategy from purely providing models to building a model + tools ecosystem, including a China-focused alternative to Cursor, the AI coding tool valued at over $2 billion in 20253

3

. While DeepSeek remains focused on advancing core model performance, rivals are shifting emphasis toward integrating AI into consumer services with more consumer appeal. Alibaba's Qwen chatbot has recently experimented with enabling users to purchase goods directly through conversational prompts via API. "The surprise would be if some of these new models end up being underwhelming. I think there are high expectations here," said Alfredo Montufar-Helu, a managing director at Ankura Consulting in Beijing2

2

. If open-source AI models keep improving, they will change where innovation happens and who sets the standards in the AI ecosystem1

1

.