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
-
AI模型在工具调用方面得到改进并修复了错误
一款新工具已被开发出来，满足了Andrej Karpathy提出的需求，据报道其开发仅用了48小时。另外，SGLang开源推理引擎中影响DeepSeek V4输出的一个错误已得到解决。此外，NousResearch的Ornstein-Hermes-3.6-27B模型的工具调用能力也得到了改进。
-
DeepSeek V4 launches with near SOTA intelligence at a fraction of the cost of competitors
DeepSeek has released its V4 model, offering intelligence comparable to leading models like GPT-5.5 and Opus 4.7 but at a significantly lower cost. This new model aims to provide near state-of-the-art performance at a f…
-
DeepSeek V4 凭借效率和华为集成挑战美国 AI 主导地位
AI 模型 DeepSeek V4 已发布，展示了性能和效率方面的显著进步。据报道，它使用的内存比以前的版本少 9.5 倍，为运营提供了可观的成本节省。此次发布挑战了美国在人工智能领域的既有优势，并突显了中国在该领域日益增长的能力，特别是其与华为 Ascend 支持框架和芯片的集成。
-
DeepSeek's new AI model promises to transform search with faster, more accurate results.
DeepSeek has developed a new AI model, DeepSeek-V4, which promises to significantly enhance search capabilities. This model is designed to deliver quicker and more precise results, potentially transforming operations ac…
-
DeepSeek V4's 1.6T parameters challenge US AI dominance with strong coding
DeepSeek V4, a new model with 1.6 trillion parameters, has been released and is showing impressive coding capabilities and efficiency. Despite facing hardware limitations in China, the model significantly narrows the pe…
-
DeepSeek V4 launches with 1.6 trillion parameters, challenging US AI dominance
DeepSeek has unveiled its V4 model, boasting an impressive 1.6 trillion parameters. This development is positioned as a significant advancement in artificial intelligence, potentially challenging the current dominance o…
-
Deepseek V4 aims for energy efficiency in AI race
DeepSeek V4, a new iteration of the AI model, is being evaluated for its potential to reduce energy consumption compared to its competitors. The model's efficiency is a key focus, suggesting a potential shift in the AI …
-
Cursor removes DeepSeek models, frustrating users seeking alternatives
Users of the Cursor IDE are reporting that DeepSeek models have been removed from the platform. One user noted the absence of DeepSeek models, which they found to be performant, and inquired about any public explanation…
-
DeepSeek-V4, LoRA, and other LLM techniques detailed in new blogs
A series of six blog posts has been published on Outcome School, detailing fundamental components of contemporary large language models. The posts cover technical concepts such as RMSNorm, DeepSeek-V4, LoRA, RoPE, GQA, …
-
美国警告全球盟友中国人工智能公司窃取知识产权
美国国务院发布了一份外交电报，警告全球合作伙伴注意中国公司（包括DeepSeek）涉嫌从美国人工智能实验室窃取知识产权。该电报强调了对“蒸馏”人工智能模型的担忧，这些模型能够以较低的成本复制功能，但可能缺乏完整的性能和安全协议。此举是在白宫和OpenAI就DeepSeek针对美国人工智能公司提出类似指控之后。
-
DeepSeek releases V4 model update for enhanced local LLM performance
DeepSeek has released an update to their V4 model, showcasing significant improvements in performance. The new version demonstrates enhanced capabilities across various benchmarks, positioning it as a strong contender i…
-
DeepSeek V4 models offer high performance with reduced inference costs and NPU support
DeepSeek has released its V4 family of open-weight large language models, featuring a 1.6 trillion parameter model and a smaller 284 billion parameter Flash MoE model. These new models claim to rival top proprietary LLM…
-
Fireworks AI launches DeepSeek V4, offering advanced inference infrastructure
Fireworks AI has announced the release of DeepSeek V4, a new large language model. The announcement was made on X, with a celebratory tone, comparing the release to a holiday event. The company is working to bring the m…
-
Deepseek V4 model rumored to achieve AGI capabilities
DeepSeek has reportedly released its V4 model, with claims of achieving AGI capabilities. The model is said to have surpassed GPT-4 on several benchmarks, including coding and reasoning tasks. This development suggests …
-
DeepSeek previews new AI model that ‘closes the gap’ with frontier models
DeepSeek has released its V4 AI model, featuring two versions: V4-Pro and V4-Flash. These models boast a 1 million token context window and utilize a mixture-of-experts architecture for efficiency. While DeepSeek V4 aim…
-
DeepSeek推出强大开源AI模型，可与顶级闭源模型媲美
DeepSeek发布了其新款DeepSeek-V4模型的预览版本，该公司声称这是最强大的开源平台，可与OpenAI和DeepMind的闭源模型相媲美。该模型已适配华为芯片技术。此次发布是人工智能日益用于增强网络犯罪能力这一更广泛趋势的一部分，使得攻击更快、更复杂。此外，该集群还触及了人工智能在医疗保健领域的日益增长的应用，指出尽管准确性正在提高，但对患者治疗结果的影响仍不清楚。
-
Meta's MCI raises privacy concerns amid AI training data harvesting
Meta is reportedly implementing mandatory data harvesting from its employees' work to train future AI models. This practice, described as potentially sinister, raises concerns about workplace surveillance and the ethica…
-
AI models generate quirky images and access GPT-5.5 via Codex backdoor
Simon Willison's blog posts highlight a humorous interaction with ChatGPT Images 2.0, which independently added a "WHY ARE YOU LIKE THIS" sign to an image of a horse riding an astronaut on a pelican riding a bicycle. Th…
-
Qwen3.6-27B model offers flagship coding performance in a smaller package
Qwen has released Qwen3.6-27B, an open-weight model that reportedly matches flagship-level coding performance. This new model significantly outperforms its predecessor, Qwen3.5-397B-A17B, while being substantially small…
-
Anthropic's Claude Opus 4.7 offers enhanced reasoning and larger context
Anthropic has released Claude Opus 4.7, a new model featuring enhanced thinking capabilities and increased token limits. This update introduces new boolean options for 'thinking_display' and 'thinking_adaptive' function…