AI: DeepSeek launches its V4 models on Huawei chips
On April 24, 2026, DeepSeek launches its V4 version, a revolutionary AI model optimized for Huawei Ascend chips. With 1.6 trillion parameters and a context of one million tokens, this open-source model challenges OpenAI and Google, marking a major breakthrough for Chinese artificial intelligence.
In brief
- DeepSeek V4 optimized for Huawei Ascend 950 chips offers 1.6 trillion parameters and reduces costs by 73% compared to the previous version.
- This open-source model outperforms all other models in reasoning and coding, closely trailing Google’s and OpenAI’s closed solutions.
- The collaboration between DeepSeek and Huawei marks a key step towards Chinese technological autonomy, challenging Nvidia.
DeepSeek V4: the AI model optimized for Huawei chips
On April 24, 2026, DeepSeek marked a historic milestone by launching the preview version of its V4 models, fully optimized for Huawei Ascend 950 chips. This breakthrough breaks dependence on Nvidia processors targeted by American restrictions. The V4-Pro (1.6 trillion parameters) and V4-Flash (284 billion) models stand out for their exceptional efficiency, reducing FLOPs by 73% and KV cache size by 90%. Major innovations include:
- A hybrid attention architecture (CSA + HCA) to handle ultra-long contexts of 1 million tokens;
- Enhanced residual connections (mHC) for increased stability.
The strategic collaboration between DeepSeek and Huawei is part of China’s quest for technological autonomy. Huawei’s Ascend supernodes will fully support the V4 models, a first for an AI model of this scale. Furthermore, benchmarks show performance comparable to GPT-5.4 and Gemini-3.1-Pro, while remaining open-source and accessible. Its capacity to manage contexts of 1 million tokens makes it an ideal tool for complex tasks such as legal analysis or academic research.
Which crypto could benefit from the DeepSeek V4 breakthrough?
The arrival of DeepSeek V4 could revolutionize crypto projects focused on AI. Here are the cryptos best positioned to take advantage of this innovation:
- Bittensor (TAO) offers a decentralized marketplace for AI models. The integration of DeepSeek V4 could significantly improve the accuracy of TAO subnetworks, especially for complex reasoning and market analysis tasks;
- Render (RNDR) provides essential decentralized GPU infrastructure to train and deploy models like DeepSeek V4. With the rise of Chinese open-source models, demand for decentralized computing power is expected to explode, especially if Nvidia chip restrictions persist;
- Fetch.ai (FET) develops autonomous agents capable of negotiating and executing smart contracts. Integrating DeepSeek V4 could enhance these agents’ decision-making, particularly for crypto trading or DeFi portfolio management.
DeepSeek V4 powered by Huawei Ascend chips represents a major advancement for artificial intelligence. This open-source, efficient, and economical model could democratize access to cutting-edge AI. For investors, Bittensor, Render, and Fetch.ai might offer opportunities. What impact do you think DeepSeek V4 could have on the tech and crypto industries?
Maximize your Cointribune experience with our "Read to Earn" program! For every article you read, earn points and access exclusive rewards. Sign up now and start earning benefits.
The world is evolving and adaptation is the best weapon to survive in this undulating universe. Originally a crypto community manager, I am interested in anything that is directly or indirectly related to blockchain and its derivatives. To share my experience and promote a field that I am passionate about, nothing is better than writing informative and relaxed articles.
The views, thoughts, and opinions expressed in this article belong solely to the author, and should not be taken as investment advice. Do your own research before taking any investment decisions.