DeepSeek

Latest: llama-modchip Runs a 284B DeepSeek V4-Flash Model in 57GB of RAM on a CPU-Only 64GB Box, Hits 9-14 Tok/s · github.com, Nov 8, 2026

Company Tracked keywords: deepseek

18 AI Weekly issues and 68 live stories from the last 30 days mention DeepSeek.

60 total mentions 2/20 recent issues -100% vs. baseline peak 2/issue tracked since 2026-03-06

DeepSeek redefined the open-weight ceiling in late April. V4-Pro (1.6T-param MoE) and V4-Flash (284B) shipped on April 24, both with 1M context, both under Apache 2.0, with quality reportedly close to Claude Opus 4.6 non-thinking and best-in-class agentic coding among open weights.

The financial story shifted simultaneously: Tencent and Alibaba are in talks to fund DeepSeek above $20B — the open-weight strategy now has its own checkbook for the first time. The technical report introduced FP4 MoE training, Compressed Sparse Attention, and the lowest KV cache footprint in any frontier release. AI Weekly tracks every DeepSeek release and the broader open-weight race.

Mention Trend — Last 20 Issues

Live Stories (68)

llama-modchip Runs a 284B DeepSeek V4-Flash Model in 57GB of RAM on a CPU-Only 64GB Box, Hits 9-14 Tok/s

github.com · 2026-11-08 · Score: 17

DeepSeek's Open-Source Harness Agent Hits 1.54M Daily Users in Under Three Months, Burns 92T Tokens a Day

bloomberg.com · 2026-11-08 · Score: 23

China's NDRC Readies Full-Lifecycle AI Safety Regulation Covering DeepSeek, Alibaba, Moonshot

bloomberg.com · 2026-11-03 · Score: 23

Rest of World: Hirundo and Lazarus Scrub Chinese Model Censorship, Deploy 'ReAligned' Qwen at ~12 US Firms

restofworld.org · 2026-11-02 · Score: 27

Researcher Finds 26,400 Working Cloud and AI Keys in Public Gitee Repos, Including OpenAI and DeepSeek

thehackernews.com · 2026-10-29 · Score: 26

The Information: DeepSeek Closes Nearly $16B Second Round Led by China's Big Fund With 27% State Capital

theinformation.com · 2026-10-29 · Score: 23

The Information: Tencent and Alibaba in Talks to Invest in DeepSeek at $20B+ Valuation

theinformation.com · 2026-10-29 · Score: 22

The Information: DeepSeek V4.1-Flash and Moonshot's Kimi K3.1 Pro Were Trained on GPT-6 Astra Distillation

theinformation.com · 2026-10-28 · Score: 23

ISTA-DASLab Ships 2-Bit DeepSeek V4.1 Flash GGUF at 100% Task Recovery on 12GB GPU

huggingface.co · 2026-10-27 · Score: 22

Artificial Analysis Intelligence Index v5 Widens Open-vs-Closed Model Gap to 12 Points

artificialanalysis.ai · 2026-10-27 · Score: 26

DeepSeek Opens ~150 Harness-Team Roles, Pitches Investors a Growth Plan Spanning Research, Infra and Ulanqab Data Centers

technode.com · 2026-10-27 · Score: 22

Cambricon Q3 Net Profit Jumps 148% to $210M on China AI Chip Push, But Revenue Misses by 44%

scmp.com · 2026-10-23 · Score: 22

China Clears DeepSeek and MiniMax as First AI Vendors Approved to Open-Source and License Models Overseas

bloomberg.com · 2026-10-23 · Score: 23

ALERT DeepSeek Raises Roughly $15B Days After US Orders Federal Agencies to Rip Out Its Tech

techcrunch.com · 2026-10-22 · Score: 36

SemiAnalysis: DeepSeek Rewrote FlashMLA KV-Cache for Huawei Ascend, Sees Full Domestic Stack by 2027-28

semianalysis.com · 2026-10-21 · Score: 17

China's MOFCOM Threatens 'All Necessary Measures' Over US DeepSeek Federal-Procurement Ban

nbcnews.com · 2026-10-21 · Score: 23

ALERT DeepSeek Open-Sources DeepInfer, Its Production Inference Framework for V4.1-Flash on Nvidia and Huawei

github.com · 2026-10-20 · Score: 34

Devs Use Claude Code and DeepSeek V4.1 Flash to Strip Ads and Trackers From Smart TVs

techcrunch.com · 2026-10-19 · Score: 22

The Information: DeepSeek Will Pre-Train on Huawei Ascend 950DT by End of 2026 in Deepening Chip Pact

theinformation.com · 2026-10-19 · Score: 23

DHS Orders Civilian Federal Agencies to Rip Out DeepSeek Tech and Bans Future Procurement

wsj.com · 2026-10-19 · Score: 23

Reuters: Chinese Hacker Breached 100+ Firms Including Deere and Cummins for $5K Using Claude Code and DeepSeek Agents

reuters.com · 2026-10-19 · Score: 23

ALERT DeepSeek Ships V4.1-Flash-1017 With Live Web Search API on 552B Agent MoE

huggingface.co · 2026-10-17 · Score: 33

ASU τ-Emotion-Bench Shows Top Agent DeepSeek V4.1-Flash Hits Only 32.2% Pass³ Under Emotional Users

huggingface.co · 2026-10-15 · Score: 21

ALERT SemiAnalysis: Just 3.6% of 857 Chinese AI Model Releases From 2021-2026 Shipped Safety Evaluations

newsletter.semianalysis.com · 2026-10-09 · Score: 28

ALERT ARTEX Developer Takes AI Pentest Tool Closed-Source After Attribution to South Korean Bank Hacks

thehackernews.com · 2026-10-08 · Score: 30

EngramEdit Hits 97% Generalization on DeepSeek Engram Knowledge Edits via Jointly Allocated N-Gram Updates

huggingface.co · 2026-10-08 · Score: 25

Nikkei: China's DeepSeek and Peers Shipped 16 AI Models in September Despite Anthropic's Slow-Down Warnings

asia.nikkei.com · 2026-10-07 · Score: 22

DeepSeek Close to $12B Pre-IPO Round With Tencent and CATL as Lead Backers, Eyes Early-2027 Listing

bloomberg.com · 2026-10-06 · Score: 23

Bloomberg Intelligence: US Lead Over China in Top AI Models Narrows to Record 3% After DeepSeek V4.1 Flash

bloomberg.com · 2026-10-04 · Score: 21

Nvidia-Backed Reflection AI Preps First Open-Weight Model to Rival DeepSeek and Qwen

axios.com · 2026-10-04 · Score: 23

ALERT OpenAI Discloses Moonshot-Linked Users Ran 16,000-Request Campaign to Extract Its Model Reasoning

thenextweb.com · 2026-09-30 · Score: 31

ALERT Reuters: 20+ Studies Show Chinese AI Agents From Alibaba, DeepSeek and Moonshot Lie in 88% of Test Sessions

reuters.com · 2026-09-29 · Score: 31

ALERT Bloomberg: China Expands Exec Travel Curbs to Spouses and Children of Top AI and Chip Talent

bloomberg.com · 2026-09-28 · Score: 31

DeepSeek Publishes DSec Paper on arXiv Detailing Elastic Compute for Model Serving

arxiv.org · 2026-09-26 · Score: 22

ALERT DeepSeek Annualized Revenue Hits $1B, Aims to Close $7.5B Shanghai Fundraise by End of October

theinformation.com · 2026-09-24 · Score: 32

China's CAC Probes DeepSeek and Moonshot Over Alleged Data Leaks to Anthropic via Claude

theinformation.com · 2026-09-22 · Score: 23

DeepSeek CEO Liang Wenfeng Calls Huawei Training Chips One of Company's Biggest Bets, Expects Q4 2026 Delivery

theinformation.com · 2026-09-21 · Score: 23

'Emergence World' Runs 8 LLM Agent Worlds for 3 Weeks, Finds Alignment Doesn't Compose in Multi-Agent Systems

huggingface.co · 2026-09-15 · Score: 27

DeepSeek Engineer Shengyu Liu: Anthropic Reaching AGI First Would Be Like Hitler Getting the Atomic Bomb

scmp.com · 2026-09-15 · Score: 26

ALERT SemiAnalysis Benchmarks Vera Rubin NVL72 at 7x Blackwell's Tokens per MW on DeepSeek V4 Pro 1.6T

semianalysis.com · 2026-09-14 · Score: 31

ALERT Russian-Speaking Actor Deploys AI Agents to Breach 395 Organizations via PaperCut Flaws in Under 6 Hours

helpnetsecurity.com · 2026-09-11 · Score: 33

ALERT Anthropic Threat Report Disrupts Bio-Weapons Plots, Chinese Distillation and Russian Cyber Ops on Claude

anthropic.com · 2026-09-10 · Score: 34

China Commerce Ministry Offers US AI Talks but Threatens Retaliation Over Distillation Claims

scmp.com · 2026-09-10 · Score: 24

DeepSeek Harness CVE-2026-82533 Let Sandboxed Agents Turn Off Their Own Sandbox With One Shell Command

thehackernews.com · 2026-09-09 · Score: 27

DeepSeek Launches 'Unprecedented' Hiring Spree for 150 Senior Engineers to Rewrite Overwhelmed Backend

scmp.com · 2026-09-08 · Score: 22

Rest of World: Chinese Banks and Telecoms Package AI Tokens Like Credit-Card Points and Data Plans

restofworld.org · 2026-09-04 · Score: 19

Kai-Fu Lee to Bloomberg: China-US AI Gap Narrows to Six Months, Race Is Now 'iPhone vs Android'

bloomberg.com · 2026-09-04 · Score: 19

DeepSeek Orders 160,000 Huawei Ascend 950DT Chips for 1GW Inner Mongolia AI Data Center

bloomberg.com · 2026-09-04 · Score: 23

Vals AI: Agentic Tasks Rack Up 10,000x the Environmental Footprint of Simple LLM Queries

bloomberg.com · 2026-09-03 · Score: 23

ALERT GreyNoise: Scanners Impersonating OpenAI, Anthropic and DeepSeek Crawlers Hunt Credentials Across 824 IPs

helpnetsecurity.com · 2026-08-31 · Score: 28

ALERT DeepSeek Drops V4-Flash-Vision-Exp, a 305B MIT-Licensed Multimodal Agent Model

huggingface.co · 2026-08-31 · Score: 31

TeamT5: Chinese State Groups Grimfengxi, Huapi and Teleboyi Turn to DeepSeek to Automate Exploits and Recon

bloomberg.com · 2026-08-24 · Score: 23

WSJ: Tsinghua Talent Engine Behind Z.ai, Moonshot and DeepSeek Shows China's AI Leap Was No Accident

wsj.com · 2026-08-22 · Score: 21

Bloomberg Feature: US-China AI Gap Is Narrowing as Cheap Chinese Open Models Get Enterprise-Grade

bloomberg.com · 2026-08-21 · Score: 20

Wired: Ulanqab, Inner Mongolia Emerges as China's AI Data-Center Capital With ~100 Sites Built or Under Construction

wired.com · 2026-08-22 · Score: 21

ALERT DeepSeek Launches V4-Flash-Vision-Exp on API, Beats Anthropic Opus 4.8 on Two Multimodal Benchmarks

siliconangle.com · 2026-08-21 · Score: 33

SemiAnalysis: Open Models Now Halve the Time to Catch the Frontier Each Cycle, Kimi K2.6 Beat Opus 4.5 in 4.8 Months

newsletter.semianalysis.com · 2026-08-21 · Score: 26

DeepSeek Opens 'Everything is a Plugin' Agent Harness, Repo Live on GitHub

github.com · 2026-08-17 · Score: 20

TechCrunch: AI Safety Tests Themselves Are Becoming a Safety Risk as Sandboxes Fail to Hold Frontier Agents

techcrunch.com · 2026-08-09 · Score: 19

DeepSeek Buys $20.8M Stake in Unitree Shanghai IPO, Signs Joint Humanoid-Brain Pact

reuters.com · 2026-08-06 · Score: 23

DeepSeek Resumes $8B Fundraise at ~$74B Valuation With Monolith in Talks

bloomberg.com · 2026-08-06 · Score: 23

CISA Adds Langflow RCE Flaw to KEV Catalog After Chinese-Speaking Actor Weaponized It via DeepSeek Agent

thehackernews.com · 2026-08-05 · Score: 26

DeepSeek Warns Users of 'Significant' Price Hike, Reversing China's AI Price War

bloomberg.com · 2026-08-06 · Score: 23

NYT: African Developers Are Increasingly Picking Chinese Open-Source AI Over Costlier US Models

nytimes.com · 2026-08-05 · Score: 22

AMD MI300X Deployment Runs DeepSeek V4-Flash at 168 tok/s Single-Stream, Ships FP8 Correctness Fixes

github.com · 2026-08-04 · Score: 20

Artificial Analysis: DeepSeek V4-Flash Runs at 3¢/Test vs $1.86 for GPT-5.6 Sol and $3.15 for Claude Fable 5

reuters.com · 2026-08-03 · Score: 23

Unit 42: Chinese Actor Wired DeepSeek Into Hermes Agent to Autonomously Attack 460+ Targets via Telegram

thehackernews.com · 2026-07-31 · Score: 30

VibeThinker-3B: WeiboAI's MIT-Licensed 3B Model Claims 94.3 on AIME 2026, Matching DeepSeek V3.2 at 1/200th the Parameters

MarkTechPost · 2026-06-19 · Score: 21

Newsletter Issues (18)

PHP2Shell: Millions of Attacks, 1.5M Servers Unpatched on Day One

2026-10-23 · 1 mentions

#528 What are companies building with AI? An Applied AI Deep Dive

2026-09-01 · 2 mentions

#524 What AI models are actually coming in the next six months?

2026-08-20 · 1 mentions

#508 The Cutting Edge, Across the Board

2026-06-28 · 3 mentions

AI Geopolitics: Washington's AI kill switch broke the alliance

2026-06-21 · 2 mentions

Deep Learning: a 744B open model just lapped the labs

2026-06-21 · 3 mentions

Machine Learning: a 744B model ran across six states on raw WAN

2026-06-21 · 4 mentions

#504 America blocked its best AI. China just raised $7.4 billion.

2026-06-17 · 7 mentions

Generative AI: DeepSeek's 75% cut rewrites the cost math

2026-05-26 · 7 mentions

Frontier Research: LLMs reach Nature as the field bets against them

2026-05-26 · 2 mentions

#490 Anthropic just had AI's biggest week of 2026

2026-05-07 · 4 mentions

Gen AI: $40B compute check, open-weight reset, software bloodbath

2026-04-25 · 2 mentions

#481 Musk wants Altman fired, Anthropic passes OpenAI, Meta goes closed

2026-04-09 · 2 mentions

#477 Jensen Huang says we've achieved AGI. The benchmarks say 0.37%.

2026-03-31 · 3 mentions

#423 Issue #423: DeepSeek: Bad for Silicon Valley, Great for You?

2025-01-30 · 8 mentions

#429 Issue #429: Google DeepMind’s new AI models

2025-03-14 · 1 mentions

#467 Issue #467: Anthropic has receipts. And nobody wants to pay for AI.

2026-02-26 · 4 mentions

#469 GPT-5.4 launches, DeepSeek V4 imminent, Qwen team implodes

2026-03-06 · 4 mentions