AI Daily · Oct 30, 2026 · Fri
ICREAT AI NEWS
Mistral releases 1-trillion-parameter multimodal model Mistral Large 4 with open weights
Mistral AI releases the open weights of its flagship multimodal model Mistral Large 4 (Le Chonk), with 1 trillion parameters, leading on FrontierCode 2.0 at 19.7% and improving Terminal-Bench 4.0 by 13 percentage points to 41%.
- Mistral Large 4 (Le Chonk) released on Hugging FaceHugging Face Blog
- vLLM supports Mistral Large 4 on release dayvLLM Blog
Model releases/updates
Latent SpaceGoogle DeepMind releases Gemini 4 Carbon, topping multiple leaderboards and opening it to cybersecurity defenders
Google DeepMind releases Gemini 4 Carbon, just two weeks after Argon, ranking first on the Artificial Analysis Intelligence Index v5 with 59 points and leading on 4 sub-evaluations including HLE-Diamond and TB-Science.
Product releases/updates
Factory BlogFirst partyFactory releases Droid 1.0 agent harness
Factory releases Droid 1.0, the largest update yet to its agent harness, enabling programmatic tool calling (Script mode) by default so the model can write a piece of JavaScript to make hundreds of tool calls in a single request.
Industry news
The Verge · AIAnthropic files S-1, pitching the AGI narrative to Wall Street
Anthropic has filed an S-1 with Nasdaq to go public, with a valuation range of $1.5 trillion to $2 trillion, and trading could begin as early as next month.
Google parent Alphabet loses $500 billion in market value in a single day as AI search summaries hit ad revenue
Alphabet's market value fell from $3.97 trillion to $3.46 trillion on Thursday, shedding more than $500 billion in a single day, with shares down nearly 13% to a six-month low. Quarterly revenue rose 25% year over year to $127.6 billion, but profit fell 25%, free cash flow was negative by more than $9 billion, and the company said AI summaries in search affected the number of sellable ads in the quarter.
Tips and opinions
OpenRouter AnnouncementsFirst partyOpenRouter tests 13 types of prompt injection attacks: Armor blocks all 73 attack emails
OpenRouter tested three layers of defenses on Claude Haiku 5.5, GPT-6 Luna, DeepSeek V4.1 Flash, and GLM 5.3 Flash using 13 self-built types of prompt injection attacks plus 12 types generated by external models.
OpenRouter AnnouncementsFirst partyOpenRouter tutorial: how to extract tables from PDFs with LLMs
OpenRouter publishes a tutorial on extracting tables from PDFs with LLMs: feed the PDF as a file input along with a strict JSON schema describing the single-row structure, and the model returns typed row data directly.
Wiz ResearchFirst partyWiz research: the rise, characteristics, and trajectory of agentic attacks
Based on incident response and threat hunting investigations, Wiz outlines how agent frameworks such as ARTEX and Hermes are used in real intrusions, including public cases such as the theft of personal information of about 67,000 people at a South Korean financial institution and a ConnectWise Cloud Backup supply chain attack that attempted logins to more than 2,250 tenants.