Mistral released Mistral Large 4 (Le Chonk), a 1T-parameter open-weight model with 49B active parameters and native multimodality
OpenAI launched the Decisions API in public beta for all developers, claiming up to 10x faster decisions than GPT-6 Luna via the Responses API
OpenAI rolled out the Meetings plugin in beta for Pro and Business users on the ChatGPT macOS desktop app, with auto notes, summaries, and next steps
Claude started working inside Google Docs, Sheets, and Slides via a sidebar, with edits made in place and approved one by one
Google released EmbeddingGemma 2 under Apache 2.0, its first natively multimodal open embedding model, unifying text, images, video, audio, and code
OpenAI published a repo of novel math solutions from an unreleased internal model, with ~4,000 problems posed and ~3 hours of ChatGPT Pro compute per result on average
Sesame voice assistant got a major update with a more expressive model, image generation, connectors, computer use, and the ability to run coding agents like Devin, Codex, and Claude Code
AgentMail launched AgentID, a Sign in with Google style login for AI agents using a verified email tied to a human owner
Ampersand raised a $15M Series A led by Bessemer, added Lauri Moore to the board, and made its platform generally available
Figure’s Hark Pro went live on web, iOS, and Android, with the $100 Pro plan made free
Google launched Nano Banana 2.1, an image model with better visual design, mask-based editing, subject consistency, and more natural-looking images, available in ai.studio
Mitra assistants gained the ability to chat for you in messaging apps and talk to other assistants via Mitra Link
Mistral Large 4 hit 6th on the Open-Weight Artificial Analysis Index and scored 82 on CyberGym-E2E
Gamma launched Gamma 5 with an agent that plans, designs, and edits decks, plus PowerPoint import/export and Notion, Slack, and HubSpot connectors
Google added native Markdown file support in Drive and Docs
ChatGPT Auto-review became free for everyone signed in with a ChatGPT account, running a background agent that checks the main agent’s actions against user intent
SpaceX sought to raise $40B, ~$10B in bank loans and $30B in investment-grade debt, led by Apollo, to buy Nvidia chips
Artificial Analysis ranked Mistral Large 4 as the most intelligent model from outside the US and China, on par with DeepSeek V4.1 Flash (max)
Anthropic said it found 5.5K+ verified vulnerabilities in April-October, while Glasswing partners found 129K+ in April-July, 33K+ of them critical or high severity
Anthropic merged its Cyber Verification Program and Project Glasswing into an expanded CVP with three tiers and access to its latest models
Mysten Labs and Google Cloud announced Verifiable Agent Arbiter, an evidence layer for enterprises to prove AI agents acted as authorized, anchored on Sui and Walrus
