AI Lab
DeepSeek
82 published events · on the wire since February 2026
Latest movement
AI Infrastructure
Ubicloud says decoding change eliminates malformed tool calls in benchmark
Ubicloud says it deployed a vLLM configuration change after testing showed tool-call schema accuracy improved from 92% to 100% on the K2VV ToolCall benchmark. The company serves Kimi and DeepSeek models and verified the result across all model versions it serves. The change enables constrained deco…
Chronological record
Every published event attached to DeepSeek, newest first.
October 2026 16 events
- Moonshot says Kimi K3.1 Pro filtering reduces exploit-solving performance
- China threatens countermeasures over DHS ban on DeepSeek products
- US orders civilian agencies to remove DeepSeek technology
- DeepSeek releases inference framework for Nvidia and Huawei hardware
- DeepSeek reportedly works with Huawei on future AI chip designs
- US bars civilian federal agencies from using DeepSeek services
- DeepSeek documents web search and page retrieval APIs for agents
- DeepSeek releases V4.1-Flash update with claimed speed and capability gains
- DeepSeek releases V4.1 Flash 1017 weights and prompt integration toolkit
- Vercel adds DeepSeek V4.1 Flash 1017 to AI Gateway
- Microsoft claims higher training efficiency for MAI-Base-2
- DeepSeek reportedly agrees to lead Winverse AI seed round
September 2026 14 events
- DeepSeek releases AI chip programming software built with Huawei
- DeepSeek releases computation and communication tools for Huawei Ascend chips
- DeepSeek researchers report sandbox system running more than 380,000 agent environments at once
- Study finds math and code training can weaken AI safety safeguards
- ClosedQuorum malware delegates attack decisions to four AI models
- Nvidia reports confidential AI inference retains more than 96% of throughput
- China investigates DeepSeek and Moonshot over alleged data leaks to Anthropic
- Xiaomi releases open-weight MiMo-V2.6 models with 1-million-token context
- AWS Strands team releases deployable agent harness
- Nvidia reports 10x MoE training gain with JAX Transformer Engine path
- Anthropic says it disrupted China-based distillation campaigns against Claude
- Anthropic alleges DeepSeek and Moonshot used fake accounts to distill Claude
August 2026 14 events
- DeepSeek publishes open weights for V4 vision model
- OpenAI reports Jalapeño inference-chip efficiency results
- DeepSeek reportedly nears $7.4 billion funding round
- Nvidia presents dynamic power-allocation approach for Rubin data centers
- OpenAI reports Jalapeno chip benchmarks against Nvidia GB300
- Researchers report wider AI use by Chinese state-linked hacking groups
- DeepSeek ends weekend peak API pricing
- DeepSeek documents experimental vision model for its API
- DeepSeek raises V4 API rates as agent-task testing shows harness gaps
- Meta releases Muse Glimmer open-weight model as Zuckerberg pushes U.S. policy shift
- Multi-tier KV cache system targets GPU inference memory bottlenecks
- Qwen3.8-Max: A New Bar for Coding and Cowork
July 2026 17 events
- DeepSeek-V4-Flash Update
- Peec AI Adds Mistral Coverage, Now Tracking 10+ LLMs Across the US, China, and Europe
- DeepSeek pauses second fundraising round after online reports of founder comments, Bloomberg says
- Etched closes $300M Series C at $10.3B valuation, TechCrunch reports
- NVIDIA says Vera Rubin NVL72 is ramping with partners and 10x tokens per megawatt vs Blackwell
- Poolside drops Laguna S 2.1, an open-weight coding model that beats rivals 10x its size
- Markets may have just experienced their second DeepSeek shock, this time thanks to a Chinese AI lab named after a Pink Floyd album
- Moonshot's Kimi K3 To Be Largest Open Model With 2.8 Trillion Parameters, Has 1M Context Window
- DeepSeek's new AI model does not wow markets in fast-changing industry
- DeepSeek is preparing for IPO filing as soon as this year
- NVIDIA claims GB300 NVL72 systems deliver up to 25x performance per watt over Hopper
- China plans limited Nvidia H200 purchases for top AI firms
Show 5 earlier events
- FirstResearch Framework Makes AI-Generated Scientific Questions Auditable
- China's DeepSeek developing its own AI chip
- NVIDIA Blackwell inference software delivers up to 5× DeepSeek V4 performance improvement in one month, reducing token costs to one-fifth of prior levels
- Etched emerges from stealth with working transformer-inference ASIC, $800M raised, $1B+ in customer contracts
- DeepSeek to launch V4 in mid-July with new peak-time API pricing
June 2026 2 events
May 2026 2 events
April 2026 9 events
- Elon Musk testifies xAI 'partly' used OpenAI models to train its own
- SenseTime releases open-source image model that runs on Chinese chips
- DeepSeek previews V4 AI model, claims parity with top US systems
- DeepSeek releases V4 Pro and V4 Flash preview models at fraction of frontier prices
- White House Accuses China of Industrial-Scale AI Model Distillation
- US accuses China of industrial-scale AI theft
- AI coding startup Factory hits $1.5B valuation with $150M funding round
- Research Shows Smaller AI Models Match Large Systems on Vulnerability Detection
- Arcee AI Releases Trinity-Large-Thinking, a 399-Billion Parameter Open Source Reasoning Model
March 2026 5 events
- Fusion's DeepSeek Moment?
- House Homeland Security publishes ICYMI on PRC AI/robotics national security risks
- Rep. Fong Delivers Opening Statement in Hearing on Threats Posed by PRC Artificial Intelligence, Robotics, and Autonomous Technologies
- House Subcommittee Announces Hearing on PRC AI, Robotics and Autonomous Tech Risks
- A "diff" tool for AI: Finding behavioral differences in new models