AI Daily Brief — 5 December 2025
Meta strikes commercial AI data deals with CNN + Fox + Le Monde + People + USA Today + 3 more — first major Western frontier-lab…
Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
Meta strikes commercial AI data deals with CNN + Fox + Le Monde + People + USA Today + 3 more — first major Western frontier-lab…
Google ships Gemini 3 Deep Think to Ultra subscribers ($250/mo) — HLE 41%, ARC-AGI-2 45.1%. AWS re:Invent Day 4: 3 frontier agents (Kiro + Security +…
Merge pull request #478 from d46/main fix(sft): supported transformers for qwen3_moe
Merge pull request #530 from turkerali/patch-1 Update README.md for blog links.
Merge pull request #547 from Speakn0w/plotcraft-benchmark upload plotcraft benchmark
Research has been core to Liquid AI from the start. Today, we’re giving that work a formal name: Liquid Labs, our team driving fundamental breakthroughs in…
AWS re:Invent Day 3: Swami Sivasubramanian agentic-AI keynote. Snowflake + Anthropic $200M strategic partnership (Claude for 12,600+ enterprises via Cortex). AWS Trainium3 UltraServers GA (3nm, 2.52…
Understanding How DeepSeek's Flagship Open-Weight Models Evolved
OpenAI researchers are testing “confessions,” a method that trains models to admit when they make mistakes or act undesirably, helping improve AI honesty, transparency, and trust…
Merge pull request #111 from zRzRzRzRzRzRzR/main Ascend A3 NPU with xLLM support
Weaviate recognized by AWS Partners in Benelux as leaders in helping customers drive innovation
Mistral ships Mistral 3 family open Apache 2.0 — 675B/41B-active MoE flagship + Ministral 3B/8B/14B + 256K context. Anthropic launches Claude for Nonprofits on GivingTuesday (75%…
A family of frontier open-source multimodal models
DeepSeek releases V3.2 + V3.2-Speciale open under MIT (685B, AIME 96.0%, IMO gold 35/42, ICPC 2nd) with DeepSeek Sparse Attention paper (O(kL) near-linear). Cyber Monday all-time…
Pinecone Dedicated Read Nodes are now in Public Preview
At Liquid AI, we believe that scientific progress accelerates when knowledge is shared openly. Today, we're pleased to release the full technical report for LFM2, our…
We evaluated AI agents' ability to exploit smart contracts using a new benchmark comprising contracts that were actually exploited. On contracts exploited after the latest knowledge…
Groq Recognized in 2025 Gartner® Cool Vendor in AI Infrastructure report
Adversarial Poetry paper goes viral: 62% jailbreak success across 25 frontier models (Gemini + DeepSeek + Mistral most permissive; GPT-5 + Claude Haiku 4.5 most resistant;…
DeepSeek releases Math-V2 — first open model with IMO gold (5/6 IMO 2025 problems, 118/120 Putnam 2024 beating top human 90). Self-verifying reasoning. AI shopping +500-700%…
Black Friday smashes records: $11.8B US online (+9.1%), $14.2B global AI-driven (Salesforce). AI traffic to US retail +805% YoY. AI conversions +38% Black Friday / +54%…
Merge pull request #1848 from Zhaohai-Li/main add evaluation reproduction code for MMMU,MathVision,ODinW-13,RealWorldQA
US Thanksgiving — frontier-lab desk dark. arXiv 'Irresponsible AI: big tech's influence on AI research and associated impacts' position paper lands. OpenAI ChatGPT-logs 20M discovery order…
SAP brings North to its Sovereign Cloud, bringing European enterprises and governments efficient, secure and powerful agentic AI.