AI News

Mistral launches Mistral Large 4 preview, a 1T-parameter open-weight model

Mistral AI opened a public preview API for Mistral Large 4 (nicknamed "le Chonk"), which it calls the strongest open-weight model from the US or Europe. The natively multimodal model has 1 trillion total parameters and 49 billion active ones, was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European data centers, and its weights are due by the end of the month. On Artificial Analysis's Intelligence Index it scores 38, ahead of GLM-5.2 but well behind Claude Opus 5.5 at 58.

MMistral AI News·1d

0325

Meta, Sierra and partners launch Personal Agent Protocol for AI agent commerce

Enterprise AI startup Sierra said on Oct. 6 that it is developing the Personal Agent Protocol (PAP) with Meta and companies including Walmart, Shopify and Stripe — an open standard governing how user-authorized personal AI agents access businesses' websites, APIs and services. Sierra co-founder Bret Taylor said the protocol is designed to handle authentication, empower consumers and give companies visibility into what personal agents do through their sites, APIs or company agents, and that it is open for anyone to implement; Meta's David Singleton compared it to email as a universal standard. The move follows Amazon blocking Meta's Muse agent from its retail site and reports of agents hitting anti-bot blocks on other retail sites, where users often cannot tell whether a block is intentional or accidental.

CCNBC Technology·21h

0428

Google releases Nano Banana 2.1 image model, halving output prices

Google released Nano Banana 2.1, an image generation and editing model built on Gemini 3.6 Flash, adding mask-based local editing and stronger subject consistency — up to 14 reference images, keeping four characters and ten objects consistent — plus Google Search grounding and minimal/medium/high thinking levels. Pricing is roughly halved versus Nano Banana 2: a 1K image drops from 6.70 to 3.36 cents and a 4K image from 15.10 to 7.56 cents, while input and text/thinking output prices rise. It is available now in AI Studio (gemini-nano-banana-2.1), the Gemini app, Search AI Mode, Flow and other surfaces, and the Nano Banana 2 API retires on October 29. The Decoder notes 2.1 beats Nano Banana Pro on some benchmarks, but Pro often still produces better images in practice.

GGoogle AI Studio·1d

0514

Claude brings direct editing to Google Docs, Sheets, and Slides

Anthropic is rolling out the Claude for Google Workspace add-on in public beta, bringing Claude into sidebars in Google Docs, Sheets, and Slides. Users can revise documents, create formulas and tables, build slides, and preview changes before applying them; separate connectors also let Claude edit Google files from a Claude conversation.

99to5Google·21h

0620

Anthropic Expands Claude Startups Program

Anthropic is expanding its Claude Startups program to deepen ties with founders and fast-growing companies. The move is part of the AI company’s broader effort to build relationships with startup customers.

CCNBC Technology·22h

0737

Anthropic expands Cyber Verification Program to three access tiers for security teams

Anthropic is expanding its Cyber Verification Program (CVP) and folding the separate Project Glasswing into it as a single offering with three tiers — Defense Access, Red Team Access and Specialized Access — giving vetted security teams reduced safety blocks on Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1. On Anthropic's CyScenarioBench, Claude Opus 5.5 was blocked on all 50 trials without CVP access, 46 of 50 under Defense Access, and none under Red Team Access, where it completed 34 of 50 — equivalent to the 67.6% success rate with no safeguards. Anthropic also says Glasswing partners found at least 129,000 verified software vulnerabilities between April and July 2026, more than 33,000 of them critical or high severity, plus 5,500 from its own open-source scanning; it calls these numbers a significant undercount.

Anthropic · News·2d

086

Lambda reportedly seeks $4B ahead of planned 2027 IPO

AI cloud provider Lambda is reportedly raising up to $4 billion at a $14.5 billion pre-money valuation, potentially its final private round before a planned 2027 IPO. Coatue Management and Blackstone are leading the round, while much of Lambda’s reported $50 billion backlog appears tied to a $35 billion Anthropic commitment. The financing highlights both strong demand for scarce GPU capacity and the debt and customer-concentration risks facing neocloud providers.

TTechCrunch AI·20h

0934

Google releases EmbeddingGemma 2, a 740M on-device multimodal embedding model

Google announced EmbeddingGemma 2 on October 6, a 740M-parameter open model built on Gemma 4 and released under Apache 2.0 that maps text, code, images, video and audio into one shared 768-dimensional space. It is modular: text/code alone needs 270M parameters, with optional vision (440M) and audio (570M) encoders up to 740M for full multimodal, and Matryoshka Representation Learning lets developers truncate vectors to 512, 256 or 128 dimensions for up to 6x storage savings. Google says it scores 78.68 versus 68.76 for the previous version on MTEB (Code) and beats some rival models twice its size; with quantization it uses about 191MB of RAM for text-only on a Pixel 11 Pro. Google also showed an experimental Mac app, AI Edge Foresight, for offline note-taking and personal knowledge retrieval.

GGoogle Developers Blog·2d

109

DeepSeek reportedly nears RMB 80 billion financing round

DeepSeek is reportedly close to securing at least RMB 80 billion in a new financing round, according to Bloomberg-cited sources, with the final amount potentially approaching RMB 100 billion. The round was initially planned at about RMB 50 billion and is being linked to preparations for a possible Shanghai STAR Market IPO in early 2027, though the transaction has not been finalized.

钛钛媒体·1d

116

Atlassian and OpenAI Expand Enterprise AI Partnership

Atlassian and OpenAI are expanding their partnership, with OpenAI frontier models powering agents across Atlassian’s platform and Rovo. The agreement also expands Atlassian’s access to models including GPT-6 Astra and the GPT-5.6 series, while connecting ChatGPT and Codex to Atlassian work data through plugins and the Teamwork Graph.

OpenAI·1d

126

OpenAI adds monitoring to halt training if models access web improperly

Reporting from the Australian parliament, Victoria Kim quotes OpenAI chief strategy officer Mr. Kwon saying that since the Medicare breach the company has added monitoring that allows staff to "immediate intervention" to stop training if its models access the internet in ways they are not supposed to. The quote was excerpted on Simon Willison's blog under AI-security tags.

SSimon Willison's Weblog·16h

136

Open source

llm-openai-decisions 0.1a0 plugin wraps OpenAI's Jev-style Decisions API

Simon Willison released llm-openai-decisions 0.1a0, an LLM plugin for OpenAI's new Decisions API announced at last week's DevDay; he had GPT-6 Astra read the new API docs and modeled the plugin on his existing llm-typesafe plugin for Jev. The new decision model, gpt-6-luna, accepts image input as well as text. According to the post, both models charge for input only and not output — OpenAI at 10 cents per million input tokens versus Jev's 4.2 cents — and both support yes/no, choice and score question types.

SSimon Willison's Weblog·17h

146

Melius raises $25M after pivoting from ad optimization to AI creative tools

Melius, an AI platform for generating advertising campaigns, images, and videos, announced $25 million in total funding, including a $20 million Series A led by CRV and a $5 million seed round led by General Catalyst. The startup says it abandoned its original performance-marketing product and rebuilt the company around tools for generating creative assets and campaigns.

TTechCrunch AI·18h

156

Vinci raises $250M Series B at $1.5B valuation for AI simulation platform

Vinci4D Inc. announced a $250 million Series B round at a $1.5 billion valuation, co-led by Advent, Temasek and Xora, with participation including AMD Ventures. Its platform is built around a physics-focused AI model that turns uploaded CAD drawings into simulations, automates meshing and uses custom GPU kernels to speed convergence; the company says it can complete simulations with hundreds of millions of degrees of freedom (DOFs) in minutes, versus the hours or days such workloads normally take. Vinci's initial focus is thermal simulation for chips—processors, memory modules and interconnects—with plans to extend to vehicles, aircraft and satellites.

SSiliconANGLE AI·18h

166

Penguin Mail launches open-source Linux email client with optional local AI

Penguin Mail is an open-source Rust email and calendar client for Linux, supporting Gmail, Microsoft, IMAP and POP3 accounts. Its optional assistant can run locally through LM Studio or Ollama, uses app tools to query mail, and asks for confirmation before sending messages or changing settings.

HHacker News · AI（100+ 分）·18h

176

Norway Prepares Temporary Ban on AI Smart Glasses

Norwegian authorities are preparing a temporary ban on AI-enhanced smart glasses in places such as parks, beaches and schools while they assess privacy risks and draft potential permanent rules. Oslo has already banned the devices in schools, and Equinor has prohibited them at its offices and offshore facilities.

TTechSpot·18h

186

Dell adds knowledge graph, semantic layer and Knowledge Agents to AI Data Platform

Dell announced extensions to its AI Data Platform, adding a Unified Semantic Layer, an Enterprise Knowledge Graph and Knowledge Agents to its Data Orchestration Engine, plus a Data Processing Engine that runs on GPUs via NVIDIA's cuDF library. The platform ties storage (PowerScale, ObjectScale and the Lightning File System) to orchestration, governance, search and GPU acceleration to cut data movement and turn enterprise data into agent-ready context. Dell infrastructure chief Arthur Lewis said the constraint is often the data, not the model.

SSiliconANGLE AI·19h

196

Gemini CLI v0.63.0 ships reliability and agent-loop fixes

Google released Gemini CLI v0.63.0 with fixes for connection-recovery progress indicators, MCP configuration errors, paused stdin, tool-output limits, temporary-directory cleanup, and authentication loops. The release also enables autonomous plan execution in non-interactive mode and includes related changelog entries for earlier preview and stable versions.

GGemini CLI Releases·20h

206