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%
- instance of DeepSeek 90%
- used by Huawei Ascend 90%
- used by 36Kr 90%
- founded Liang Wenfeng 90%
- used by Moore Threads 90%
- instance of DeepSeek V4-Flash 90%
- developed by DeepSeek V4-Flash 90%
- instance of DeepSeek V3.2 90%
- instance of mixture of experts 90%
- competes with Moonshot AI 80%
- 2026-06-08 product_launch DeepSeek's V4 model launch has led to significant price cuts by Chinese cloud providers and competitors like Xiaomi. source
- 2026-06-01 research_milestone DeepSeek V4, co-designed with Huawei hardware, shows significant performance gains, indicating a shift in global AI leadership. source
- 2026-05-31 research_milestone DeepSeek V4 demonstrates strong performance in Chinese cultural contexts and legal accuracy, despite mixed global benchmark results. source
- 2026-05-25 research_milestone DeepSeek V4 completes full adaptation to Huawei Ascend chips, marking a milestone for China's domestic AI stack. source
- 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. source
- 2026-05-18 product_launch DeepSeek launched its V4 models, V4-Pro and V4-Flash, on April 24, 2026.
- 2026-05-16 product_launch DeepSeek V4, a new open-weight LLM family, was released with significant architectural improvements and cost reductions. source
- 2026-05-16 research_milestone DeepSeek V4 achieves a 98% reduction in KV-cache memory usage with its new compressed attention architecture. source
- 2026-05-15 product_launch DeepSeek released its V4 model with MegaMoE optimizations. source
- 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. source
25 day(s) with sentiment data
-
DeepSeek releases V4, the largest open-source AI model, built on Chinese chips
DeepSeek has released DeepSeek V4, an open-source AI model that is reportedly the largest ever built. This new model is capable of running on Chinese-made chips and is significantly cheaper to operate than comparable mo…
-
DeepSeek V4 lags US models by 8 months; OpenAI criticized over data collection tactics
A recent evaluation suggests that DeepSeek V4 lags behind leading US models by approximately eight months, according to NIST's assessment. This finding highlights the competitive landscape and performance gap of Chinese…
-
Simon Willison integrates iNaturalist wildlife photos into his blog using Claude Code
Simon Willison has integrated his wildlife photography into his blog, showcasing iNaturalist observations. He utilized Claude Code for web to build this feature entirely on his phone, extending his existing content synd…
-
GeoContra framework enhances LLM-driven GIS analysis with verifiable geographic rules
Researchers have developed GeoContra, a framework designed to improve the reliability of LLM-generated code for geospatial analysis. GeoContra enforces geographic rules such as coordinate semantics, topology, and plausi…
-
DeepSeek's 200-person team embarrasses AI giants with open-sourced, high-performance model
A Chinese AI team named DeepSeek has released DeepSeek V4, a 1.6 trillion parameter model with a 1 million token context window that reportedly outperforms leading models from major AI labs. Despite having a significant…
-
Huawei integrates DeepSeek V4 AI into Celia Claw for enhanced self-evolution
Huawei has integrated the DeepSeek V4 AI model into its Celia Claw virtual assistant, enhancing its capabilities. This update allows Celia Claw to learn user habits, remember preferences, and adapt its interactions for …
-
Moore Threads completes full-link engineering adaptation for DeepSeek-V4
Moore Threads has successfully adapted the DeepSeek-V4 large language model to run on its flagship AI training and inference accelerator card, the MTT S5000. This integration was achieved using the company's proprietary…
-
Simon Willison advocates for RSS feeds to share personalized, rapidly developed apps
Simon Willison is advocating for an RSS feed specifically for "vibe-coded" applications, inspired by Matt Webb's desire for an installable feed of such tools. Willison suggests that as "vibe-coding" speeds up app creati…
-
DeepSeek rolls out new vision model with fast, but sometimes flawed, image understanding
DeepSeek has begun a limited release of its new "Vision" multimodal model, which appears to be a separate entity from its V4 text-based models. Early testing reveals that while the Vision model is exceptionally fast in …
-
Hugging Face integrates DeepInfra for serverless AI model inference
Hugging Face has integrated DeepInfra as a new serverless inference provider on its Hub. This collaboration allows developers to access a wide array of models, including LLMs like DeepSeek V4 and Kimi-K2.6, through Hugg…
-
AI coding assistants should help professional software companies build better products
Matthew Yglesias, writing in April 2026, expressed a desire for AI coding assistance to be integrated into professionally managed software companies. He believes these companies should leverage AI to produce more, bette…
-
Open-source AI models surge, while a private 20T-parameter model hints at future scale
Open-source AI models are demonstrating significant performance improvements, with DeepSeek V4 and Qwen 3.6 showing capabilities that rival those of large corporate-backed models. This advancement increases the practica…
-
Alibaba's HappyHorse video model enters beta, enhancing Wukong platform
Alibaba's ATH division has launched a closed beta for its new video generation model, HappyHorse 1.0. This model, integrated into Alibaba's Wukong platform, supports text-to-video, image-to-video, and video editing, cap…
-
DeepSeek V4 First Release Adaptation Behind: Why does Ascend insist on not doing a CUDA compatibility layer?
Huawei's Ascend AI accelerators are forging a unique path by eschewing CUDA compatibility to build an independent ecosystem. This strategy focuses on deep architectural changes in their latest Ascend 950 chips to addres…
-
DeepSeek V4 prioritizes batch invariance, sacrificing GPU efficiency for stability
DeepSeek V4's technical report reveals a core design choice of "batch invariance" to ensure consistent outputs across different batch configurations and processing pipelines. This feature is crucial for maintaining repr…
-
Simon Willison releases llm-echo 0.5a0 for testing LLM tools
Simon Willison has released version 0.5a0 of llm-echo, a plugin for the LLM tool that provides a "fake" model for testing purposes. This new version allows users to test against LLM versions 0.32a0 and higher, and it ou…
-
DeepSeek V4 release sparks surge in Chinese semiconductor stocks, boosting domestic AI computing power
DeepSeek V4's release has significantly boosted China's A-share semiconductor market, with sectors like GPU and semiconductor equipment experiencing a surge. This rally is attributed to V4's compatibility with Huawei's …
-
DeepSeek slashes API prices by 90% after V4 model release
DeepSeek has significantly reduced its API prices by up to 90% following the release of its V4 model. The company attributes these price cuts, which establish a new industry low, to its sparse attention architecture. Th…
-
DeepSeek V4 launches with 1M context window and open weights
DeepSeek has released its V4 model, featuring a 1 million token context window and open weights. This release positions DeepSeek V4 as a significant advancement in the field, particularly for its accessibility through o…
-
Honor integrates DeepSeek-V4 AI model into YOYO assistant for enhanced on-device capabilities
HONOR has integrated the DeepSeek-V4 AI model into its YOYO virtual assistant, enhancing on-device capabilities. This integration makes YOYO the first Android AI agent to support the DeepSeek V4 large language model. Th…