Home/Latest News/DeepSeek News
TOPIC HUB
DeepSeek News
DeepSeek news from AiLookout: the latest DeepSeek model releases, open weights, API pricing, benchmarks, chip questions and security research, sourced.
Newest first, each with its sources linked.
About our DeepSeek coverage
DeepSeek is a Chinese AI lab known for releasing capable models with open weights and low API prices. This hub collects AiLookout’s DeepSeek news, newest first: model and agent releases, open-weight licenses, API pricing, benchmark results, questions about the chips used to train its models, government restrictions, and security research that involves DeepSeek models.
Benchmark figures published by DeepSeek are labelled as DeepSeek’s results and linked to its model card or paper. Independent evaluations, export-control reporting and security findings are linked to their own sources, so each claim can be checked.
Use this page to follow DeepSeek news and the wider open-weight model race. For other open-weight developers, see our Mistral news hub, and for closed frontier models, see OpenAI news and Claude news. Every story here links its primary sources, shows its publication date, and discloses the AI assistance used to write it. New stories are added as they are published, and older coverage stays available further down the list.
Latest DeepSeek News
Tools & Agents5 min read
DeepSeek open-sources Super-Agent, a coding agent that reads team chats
The MIT-licensed research prototype listens to GitHub and Feishu, dispatches work to isolated worker sessions, and replies in the original conversation with a pull request.
News
Security5 min read
China-linked VAPOR PANDA hackers use DeepSeek and ARTEX to attack more than 200 organizations
CrowdStrike reports China-linked VAPOR PANDA ran the open-source ARTEX agent framework on DeepSeek to attack more than 200 organizations in at least 14 countries over 11 days, with a single controller directing multi-agent intrusions from a human operator issuing instructions in Chinese.
News
Industry7 min read
DeepSeek and MiniMax clear China's new AI export registration for open-source releases
A CCTV-affiliated account said both labs completed registration under Announcement No. 48, and MiniMax confirmed it to the Hong Kong exchange on 23 October.
News
Industry8 min read
DeepSeek raises about $15 billion as CATL and Tencent lead its second round, reports say
Bloomberg and Reuters report DeepSeek has raised about 100 billion yuan ($15 billion), double its original target, with CATL and Tencent among the biggest backers, as it prepares a possible Shanghai STAR Market IPO.
News
Tools & Agents9 min read
Ant Group and vLLM launch Humming, a kernel library for quantized inference
The Apache-2.0 library compiles quantized GEMM kernels on demand for every NVIDIA GPU since Turing; its authors report up to 43% more DeepSeek-V4.1-Flash throughput on eight H20s.
News
Tools & Agents8 min read
DeepSeek open-sources DeepInfer, the engine behind its V4.1 Flash service
DeepSeek released DeepInfer under Apache 2.0 with support for NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs, plus a detailed Q&A on how it serves DeepSeek-V4.1-Flash, including a 155 trillion tokens per day cluster figure.
News
Industry8 min read
DHS orders federal civilian agencies to remove DeepSeek, including tools built on it
A FASCSA order from Homeland Security Secretary Markwayne Mullin bars DeepSeek from civilian agency purchasing and requires removal of DeepSeek products and third-party services built on DeepSeek. The order is dated 24 July 2026; its Federal Register notice appeared on 19 October.
News
Tools & Agents7 min read
Exposed Cairn panels show one hacker's AI agents breaching more than 100 company sites
NorForge watched more than 20 open Cairn control panels where AI agents probed e-commerce sites. Reuters matched Deere, SharkNinja and Cummins among the targets. Successful breaches took about a day and, by one estimate, roughly $50 in tokens.
News
Industry8 min read
Open-weight models handle 71.5% of Vercel AI Gateway tokens and 23% of spend
Vercel's October AI Gateway Production Index says open-weight models' share of spend quadrupled since June, DeepSeek carried 58% of tokens, and most spend on Chinese models went to US hosts.
News
Models8 min read
DeepSeek ships V4.1-Flash-1017 with open weights and faster, cheaper agent runs
DeepSeek replaced the model behind deepseek-flash with a 17 October snapshot that it says scores higher on all eleven reported benchmarks while finishing tasks faster with shorter outputs. Prices are unchanged, and MIT-licensed weights are on Hugging Face.
News
Tools & Agents8 min read
DeepSeek launches Search and Fetch APIs for AI agents at $0.75 per 1,000 calls
DeepSeek's new Search and Fetch endpoints give agents live web results and pages as Markdown, using its in-house search engine, billed per successful request and capped at 5 requests per second.
News
Industry10 min read
China proposes export controls on 23 AI technologies, from large models to agents
China's commerce and science ministries published a draft on 16 October 2026 that adds 23 AI technologies to the country's export-control catalogue, including weights of models above 10^26 FLOPs or 2 trillion parameters, high-performing agents and AI-driven malware. API, app and SaaS access stay outside, and open-sourced work is exempt.
News
Research9 min read
Study finds LLMs keep reasoning even with thinking mode switched off
Researchers from Tsinghua, Oxford and Stanford tested six models, including DeepSeek-V4-Flash, GPT-5.4 and Gemini 3 Flash, and found that no-think switches and “answer directly” prompts rarely stop visible reasoning, especially on open-ended questions.
Research
Tools & Agents6 min read
DeepSeek Harness v0.2 ships macOS and Windows desktop builds
DeepSeek’s Harness v0.2 release candidates add macOS and Windows desktop installers for the MIT-licensed agent runtime. The product page and SAFETY.md still call it a developer preview that can run model-generated commands.
News
Industry7 min read
CrowdStrike: ARTEX plus LLMs used against South Korean finance. Actor unnamed
A 7 October CrowdStrike Intelligence note says an unattributed, likely Chinese-speaking and financially motivated actor used the open-source ARTEX pentest agent with DeepSeek v4.1-flash, GLM-5.3, Grok 4.6 and Claude Code from late September into early October. Identity details in a résumé prompt are not confirmed. Autumn-27 later said the project is closed source.
News