Here are today's news for 11th of November about Tencent, NVIDIA, Anthropic and Google DeepMind. Let's dive into the details. On November 4, 2026, Tencent announced Hy4, the full release of its latest large language model, adding image understanding and computer, use capabilities for the first time. The model can analyze detailed images, charts and visual math problems, and work through graphical and command, line interfaces; Tencent says it was trained to use specialist software including Blender and CAD. Its August 28 Hy4 Preview had 770 billion total parameters, 49 billion active per token and a context window exceeding one million tokens. Tencent has not confirmed that those figures apply unchanged to the final release. The launch matters because Tencent is seeking a stronger place in the open, model competition with DeepSeek, Moonshot AI and Z.ai, while positioning Hy4 for practical workplace agents rather than chat alone. In Tencent’s blind financial evaluation, 213 professionals assessed 579 real, world tasks; Hy4 scored 3.04, ahead of DeepSeek, V4.1, Flash, 1017 at 2.97 and Kimi K3.1 Pro and GLM, 5.3 at 2.68 each. Tencent also reported a 12. On November 10, 2026, at KubeCon + CloudNativeCon North America, NVIDIA contributed OpenShell to the Cloud Native Computing Foundation as a Sandbox project. The open, source runtime governs autonomous AI agents’ execution, resource access and inference routing; CNCF stewardship puts its development under a vendor, neutral community rather than NVIDIA alone. That matters as companies deploy agents able to use tools and reach internal systems: an agent’s mistake or manipulation can expose data, credentials or services, making enforceable controls relevant to cloud providers, software vendors and enterprise security teams. OpenShell runs each agent in an isolated, unprivileged sandbox with kernel, level limits on files and system calls; network requests pass through a separate supervisor that applies policy and supplies real credentials only for approved endpoints. A gateway manages sandbox lifecycles and policies, while a formal policy prover checks proposed access changes. NVIDIA had introduced OpenShell 0.1. On Nov. 10, 2026, Anthropic introduced Claude Meetings, Whiteboards and Sheets, extending Claude from a personal assistant into shared workplace tools. Meetings, in beta for Team and Enterprise customers, joins Google Meet as a visible participant: it can answer in chat, retrieve information from connected work apps, and draft plans and timelines during discussion; notes and action items appear in Claude afterward. Whiteboards and Sheets launched in beta on paid plans, across web, desktop and mobile. Teams can turn brainstorms into organized boards, or create and edit multi, tab spreadsheets with formulas, formatting and dropdowns; Claude outlines cell edits, and boards can become documents or slide decks. The announcement puts Anthropic more directly against office and meeting, software providers by targeting collaborative work, not just individual AI assistance. It builds on September’s merger of Cowork and chat and launches of Claude Docs and Slides, plus April’s Claude Design. On November 10, Anthropic launched its IPO roadshow, offering 825 million Class A shares at $70, $78 each and seeking up to $64.4 billion; a concurrent $26 billion placement would lift gross proceeds to $90.4 billion. Anthropic is selling about 801.4 million shares, while existing investors offer 23.6 million. Google, Nvidia, AMD, Broadcom, Micron, Microsoft and TSMC are the private, placement buyers. The proposed Nasdaq symbol is ANT, with trading expected November 19. The deal would value the Claude maker at up to about $1.4 trillion, making this a major public, market test of investor appetite for frontier AI, and of whether explosive growth can justify the enormous cost of computing capacity. The filing reported $33.2 billion revenue for the first nine months of 2026 and an $81 billion annualized revenue run rate on October 31. Claude and coding agent Claude Code have driven demand, especially from businesses. On November 10, 2026, Google DeepMind announced that its AlphaProof Code framework had helped formally verify key security properties of libsignal, the code implementing the Signal Protocol. The protocol protects conversations in Signal, WhatsApp and Google Messages, putting the work’s potential reach at billions of users. Gemini Carbon agents generated more than six million lines of Lean proof code in a four, week project; human specialists defined and checked the specifications, and the Lean kernel mechanically checked the proofs. The work covered the PQXDH post, quantum handshake and Triple Ratchet, including authentication, handshake secrecy, forward secrecy and protection after a device is compromised and recovered. Researchers also tested the framework on older, flawed Signal versions, where it flagged the problems; they found no new vulnerabilities in the code examined. That's all for today's AI news briefing. Have a great day!