DeepSeek V4 and the AI Race: China Surpasses US Open-Source Models in Code, Gains Traction in Africa
|2 min
DeepSeek V4, with 1.6 trillion parameters and backing from Huawei, surpasses all open-source models in code generation, according to Vals AI (NYT). Its adoption is accelerating in East and West Africa.
April 24, DeepSeek launched V4 with 1.6 trillion parameters and the material support of Huawei ([see our analysis](/fr/sujets/deepseek-v4-huawei-ia-guerre-technologique-20260424)). The model represents a qualitative leap in the AI race between China and the United States.
According to independent benchmarking firm Vals AI, cited by the New York Times, 'DeepSeek V4 has significantly outperformed all other open-source systems in code generation' (source : NYT, April 24, 2026). This result is significant for two reasons: code generation is the most directly related benchmark to economic productivity, and DeepSeek V4 is open-source – accessible for free, unlike the closed models of OpenAI and Google DeepMind.
The closest historical precedent is the open-source wave of 2018-2019, when TensorFlow (Google) and PyTorch (Meta) democratized AI development by making tools free (source : Bloomberg, April 24, 2026). DeepSeek V4 replicates this schema but with a dominant Chinese actor – and material support from Huawei, under US sanctions since 2019.
African Deployment
The Least Covered Signal: DeepSeek Takes Root in African Tech Ecosystems. In Nairobi, Lagos, and Cape Town, the model offers more affordable and energy-efficient AI solutions, powering linguistic models for African languages and local startups (source: Wikipedia/DeepSeek).
Coverage is concentrated in Nigeria (Premium Times), South Africa (Daily Maverick), and Kenya (Daily Nation). American, Chinese, and Japanese media have not covered the African adoption. The Chinese Global Times presents DeepSeek V4 as a technological feat without mentioning Africa.
If DeepSeek V4 becomes the open-source standard in emerging markets, American companies will lose a strategic advantage in countries where the next billion AI users will be concentrated. The competition for AI is no longer limited to Silicon Valley and Shenzhen laboratories – it's playing out in Nairobi startups.
The Regulatory Framework Falls Behind
The US regulatory framework published on March 20, 2026 prioritizes open-weight models and online safety (source: White House AI policy framework, Gerrish Legal). The AI Foundation Model Transparency Act of 2023 defines technical criteria for foundation models (source: Wikipedia/Foundation model). However, neither addresses the scenario in which a Chinese model dominates benchmarks and gains a foothold in emerging markets. The US AI policy is designed to regulate OpenAI – not to compete with DeepSeek in Africa.
SHARE THIS COMPARISON
Show your friends how the world sees the same news differently.
AI-generated content — Analyses are produced by artificial intelligence from press articles. They may contain errors or biases. Learn more