DeepSeek V4
PulseAugur coverage of DeepSeek V4 — every cluster mentioning DeepSeek V4 across labs, papers, and developer communities, ranked by signal.
- developed by DeepSeek 100%
- subsidiary of DeepSeek 100%
- used by 36Kr 90%
- used by Huawei Ascend 90%
- used by Moore Threads 90%
- developed by DeepSeek V4-Flash 90%
- instance of DeepSeek V4-Flash 90%
- founded Liang Wenfeng 90%
- competes with Moonshot AI 80%
- used by SGLang 80%
- instance of DeepSeek 70%
- used by GitHub 70%
- 2026-05-25 research_milestone DeepSeek V4 completes full adaptation to Huawei Ascend chips, marking a milestone for China's domestic AI stack. 来源
- 2026-05-24 research_milestone DeepSeek V4 is presented as a new AI model challenging OpenAI.
- 2026-05-23 product_launch DeepSeek released its V4 model with a 1 million token context window. 来源
- 2026-05-18 product_launch DeepSeek launched its V4 models, V4-Pro and V4-Flash, on April 24, 2026.
- 2026-05-16 research_milestone DeepSeek V4 achieves a 98% reduction in KV-cache memory usage with its new compressed attention architecture. 来源
- 2026-05-16 product_launch DeepSeek V4, a new open-weight LLM family, was released with significant architectural improvements and cost reductions. 来源
- 2026-05-15 product_launch DeepSeek released its V4 model with MegaMoE optimizations. 来源
- 2026-05-11 product_launch DeepSeek V4, an AI model with a sparse Mixture-of-Experts architecture, was released.
- 2026-05-11 product_launch DeepSeek officially released its new flagship model, DeepSeek-V4. 来源
13 天有情绪数据
-
US VP Vance raises concerns over AI models like Anthropic's Mythos
US Vice President Vance has expressed concerns regarding advanced AI models like Anthropic's Mythos. These models, capable of autonomously identifying software vulnerabilities, pose a significant risk of cyberattacks on…
-
Trump plans to invite Nvidia, Apple, Exxon CEOs on China visit
Donald Trump is reportedly planning to invite CEOs from major US companies, including Nvidia, Apple, ExxonMobil, and Boeing, to accompany him on a potential visit to China. The selection process for these invitations is…
-
DeepSeek-V4 powers GitHub's top trending Agent project
A developer has created an agent using DeepSeek-V4, which has gained significant traction on GitHub, topping the trending list. This development highlights the growing interest and capability in building autonomous AI agents.
-
China grants 6G test frequencies, DeepSeek-V4 powers GitHub's top agent project
China's Ministry of Industry and Information Technology has granted the IMT-2030 (6G) Promotion Group permission to use the 6GHz frequency band for 6G technology trials. This move aims to accelerate the research, standa…
-
EU warns airlines must compensate passengers for fuel price-driven flight cancellations
The European Union has warned airlines that they must compensate passengers for flights canceled due to rising fuel prices. This directive comes as Middle East conflicts have increased aviation fuel costs, leading sever…
-
DeepSeek-V4 powers GitHub trending Agent project, while Japan sees wage growth
A developer has created an agent using the DeepSeek-V4 model, which has gained significant traction on GitHub. This development highlights the growing interest and capability in building autonomous AI agents. The succes…
-
OpenAI's custom AI chip deal with Broadcom hits funding snag
OpenAI's collaboration with Broadcom on custom AI chips is facing financial hurdles, with Broadcom reportedly seeking a commitment from Microsoft to purchase 40% of the chip production capacity. The project, codenamed "…
-
DeepSeek V4 launches on Huawei chips, undercutting GPT-5.5 by 97%
China has developed the DeepSeek V4 AI model, which is reportedly 97% cheaper than OpenAI's GPT-5.5. This development comes despite US export bans on chips intended to prevent China from advancing its AI capabilities. T…
-
DeepSeek V4 AI model offers free, high-performance alternative to costly systems
DeepSeek V4, an open-source large language model, has demonstrated performance competitive with proprietary systems costing billions to develop. The model achieves state-of-the-art results on several benchmarks, includi…
-
DeepSeek TUI coding agent gains 8,700+ GitHub stars in days
DeepSeek TUI, a new AI coding agent developed in Rust, has rapidly gained popularity, achieving over 8,700 GitHub stars within days of its release. This tool is designed to work with the DeepSeek V4 model and offers a n…
-
Heretic 1.3 发布，本地AI模型大幅降低成本，苹果就Siri AI索赔达成和解
Heretic 1.3 已发布，引入了可复现的模型输出和用于验证去审查LLM的集成基准测试系统。此次更新还侧重于减少VRAM使用量并扩展对各种模型支持。另外，一位开发者的实验表明，像DeepSeek V4和Qwen 3.6 27B这样的本地AI模型在编码任务中可以媲美云API的性能，潜在地将云成本降低高达74%或94%。在其他新闻中，苹果公司已同意支付2.5亿美元和解金，以解决关于Siri AI功能的误导性声明，符合条件的美国iPho…
-
Cloudflare cuts 1,100 jobs, citing AI efficiency gains
Cloudflare is laying off over 1,100 employees, approximately 20% of its workforce, as it transitions to an AI-first operating model. The company's CEO, Matthew Prince, stated that AI-driven automation has reduced the ne…
-
DeepSeek-V4's million-token context window shifts agent focus from fitting to retrieval
DeepSeek-V4 has been released with a million-token context window, a significant increase that shifts the challenges in building AI agents. While this long context removes the previous constraint of fitting information …
-
DeepSeek V4's architecture slashes costs, impresses with high coding benchmark performance
DeepSeek V4 was released on April 24, 2026, with its architecture featuring five key tricks that contribute to its cost-effectiveness. The model achieved a Codeforces rating of 3,206, marking it as the highest rating ev…
-
Y Combinator's 0.6% stake in OpenAI valued at over $5 billion
Y Combinator holds a stake of approximately 0.6% in OpenAI, which is valued at over $5 billion given OpenAI's current $852 billion valuation. This information was reportedly shared by an individual familiar with OpenAI'…
-
Andy Masley argues data center land sales don't threaten farmland
Andy Masley, a farmer from Loudoun County, has criticized the argument that data center construction is consuming excessive farmland. He points out that the total land sold by farmers for data centers between 2000 and 2…
-
Redis adds new array data type with AI-assisted development and browser playground
Salvatore Sanfilippo has submitted a pull request to add a new array data type to Redis, introducing numerous commands for array manipulation and querying. An interactive playground has been developed using Claude Code …
-
Born ML 框架通过 v0.8.0 更新实现純 Go GPU 加速
Born ML 发布了 0.8.0 版本，将其 GPU 后端从 Rust 的 wgpu-native 迁移到一个純 Go 实现 gogpu/wgpu。这一改变消除了对 .dll 或 .so 文件等外部运行时依赖的需求，允许 GPU 加速的机器学习应用程序进行单一二进制文件部署。此次更新还包括在迁移过程中发现的五个关键 GPU 错误的修复，并在实际模型训练中进行了验证。
-
New mHC architecture in AI models alters attention head behavior
Researchers have investigated the impact of Manifold-Constrained Hyper-Connections (mHC), a novel architecture implemented in Deepseek v4, on model interpretability. Experiments revealed that previous token attention he…
-
DeepSeek V4, Kimi K2.6 challenge top AI models; chatbot ID bill advances
DeepSeek V4 and Kimi K2.6 are emerging as strong contenders, offering competitive benchmarks and pricing that challenge established frontier AI labs. Meanwhile, a legislative effort in the U.S. to require identity verif…