{"data":{"id":1333,"lang":"en","title":"DeepSeek V4 Becomes First Frontier AI Model Built on Chinese Huawei Chips, Challenging US Sanctions Strategy","summary":"DeepSeek V4, a trillion-parameter open-source AI model, will be the first to run on Huawei chips instead of Nvidia, trained for just $5.2 million.","category":"technology","category_label":"Technology","published_at":"2026-04-07T11:03:14Z","updated_at":"2026-04-07T11:03:14Z","url":"https://techtools.cz/tools/news/article/1333-deepseek-v4-becomes-first-frontier-ai-model-built-on-chinese-huawei-chips","api_url":"https://techtools.cz/news-api/v1/articles/1333?lang=en","markdown_url":"https://techtools.cz/news-api/v1/articles/1333.md?lang=en","image_url":"https://help.apiyi.com/wp-content/uploads/2026/04/deepseek-v4-1t-moe-multimodal-april-release-guide-en-image-0.png","word_count":161,"reading_time_minutes":1,"sources":[{"url":"https://findskill.ai/blog/deepseek-v4-release-date-specs/","domain":"findskill.ai"},{"url":"https://www.nxcode.io/resources/news/deepseek-v4-release-specs-benchmarks-2026","domain":"nxcode.io"}],"translation":{"lang":"cs","id":1391,"title":"DeepSeek V4 se stává prvním špičkovým AI modelem postaveným na čínských čipech Huawei a zpochybňuje americké sankce","url":"https://techtools.cz/tools/news/clanek/1391-deepseek-v4-se-stava-prvnim-spickovym-ai-modelem-postavenym-na-cinskych-cipech","api_url":"https://techtools.cz/news-api/v1/articles/1391?lang=cs"},"content":"Chinese AI lab DeepSeek is preparing to release V4, a one-trillion-parameter open-source model that will run on Huawei's Ascend chips instead of Nvidia hardware. Reuters confirmed on April 4 that this makes DeepSeek V4 the first frontier AI model built entirely on Chinese semiconductor infrastructure, directly challenging the assumptions underlying American chip export restrictions.\n\nThe model uses a Mixture-of-Experts architecture with approximately 37 billion active parameters per token and features a one-million-token context window powered by a novel Engram Conditional Memory system. Early benchmark leaks suggest SWE-bench scores above 80% and HumanEval at 90%, putting it competitive with leading Western models. Training reportedly cost just $5.2 million — a fraction of the $100 million-plus budgets typical at this scale.\n\nDeepSeek plans to release V4 under an Apache 2.0 license with API pricing around $0.28 per million tokens, dramatically undercutting closed-source competitors. The combination of frontier-level performance, open weights, and independence from Western chips marks a significant shift in the global AI landscape."}}