Here are today's news for 21st of November about Anthropic, Microsoft, Google and OpenAI. Let's dive into the details. Today in brief: Anthropic priced its IPO at $83 a share on November 18, raising a record $81.3 billion; Microsoft announced seven in, house AI models at Ignite, including MAI Code 2 for GitHub Copilot and new image, voice, and cybersecurity models; Google opened its Genie 4 world model to developers, enabling them to create and interact with generated 720p worlds through its API; OpenAI released GPT, Live, 1.1 in its API, a voice model designed to listen and speak simultaneously and handle interruptions and OpenAI launched MedChemBench, a benchmark of 99 medicinal, chemistry problems to assess AI models’ drug, discovery reasoning. Anthropic priced its IPO at $83 a share on November 18, 2026, above its marketed $70, $78 range, selling 980 million shares and raising $81.3 billion. The total included 956.35 million shares sold by Anthropic and 23.65 million by existing holders; the offering valued the Claude maker at about $1.5 trillion. Its record haul surpassed SpaceX’s $75 billion IPO earlier this year, making Anthropic the first major frontier, AI developer to enter public markets. Strong demand prompted the company to expand the planned sale from 825 million shares. The debut gives public investors direct exposure to a business whose costly model development and computing needs require vast capital, while raising the stakes for rivals such as OpenAI and Google. Founded in 2021 by former OpenAI employees, including Dario and Daniela Amodei, Anthropic sells Claude models and services to businesses. It reported $33.2 billion in revenue for the first nine months of 2026, but also disclosed nearly $901 billion in computing and infrastructure commitments. At Microsoft Ignite on November 17, Microsoft unveiled seven in, house AI models spanning coding, cybersecurity, image generation, speech and decision, making. MAI Code 2 and the lighter Code 2 Flash are available in GitHub Copilot; Flash is also in limited preview on Microsoft Foundry. The lineup includes MAI Cyber 2 Flash, MAI Image 2.7, MAI Realtime 1 and Dragon Copilot Transcribe. Microsoft says Code 2 averages 98% of GPT, 6.1 Sol’s performance at 40% of its price, and 96% of Claude Opus 5.5’s at 5% of the cost, across DeepSWE 1.1, Terminal Bench 2.1, SWE, Atlas and FrontierCode v1.1. Those are company, reported comparisons, not independent guarantees. They signal Microsoft’s effort to make its products less dependent on outside models while giving developers cheaper coding options. Cyber 2 Flash also targets security work, illustrating how the company is building specialized models for its own software and cloud services. On Nov. 19, 2026, Google made its Genie 4 world model generally available to developers through the Gemini API and Google AI Studio, a day after starting its Project Genie rollout for eligible Google AI Ultra subscribers. Via the Interactions API, developers can generate continuous 1280×720 worlds at 24 frames per second from text, a 16:9 reference image or both, then steer scenes with keyboard controls and live prompts. Streams include synchronized 48 kHz stereo audio, from environmental sound to dialogue; sessions can last up to 120 seconds. Access requires a paid API tier: Google lists video output at $17.50 per million tokens, about $0.10 per second with audio included, and input at $1.50 per million tokens. The launch gives app and game makers a way to embed responsive, generated environments rather than confining users to Google’s own experience, intensifying competition in interactive AI and simulation. On November 19, 2026, OpenAI added GPT, Live, 1.1 to its API for real, time, full, duplex voice conversations: it can listen while speaking, handle interruptions and hand complex reasoning or tool tasks to a developer, selected backend. The model improves multilingual transcription, pronunciation, accent consistency and instruction, following over GPT, Live, 1. Developers can use the ID `gpt, live, 1.1`; active voice sessions cost $0.05 per minute. A new standby setting pauses inference after 5, 60 seconds of silence while keeping WebRTC open, then resumes when the caller speaks; standby costs $0.02 per minute, with backend models and tools charged separately. The release matters to businesses building voice agents, which can reduce idle, session costs without forcing users to restart conversations, while choosing their own reasoning systems. It also extends a growing contest over natural, low, latency voice interfaces. GPT, Live, 1. On Nov. 20, 2026, OpenAI introduced MedChemBench, a 99, problem test of whether AI can interpret molecular structures and experimental results, compare compounds, and predict their properties. The problems draw on 117 papers and patents and cover 761 distinct molecular structures. At the xhigh reasoning setting, GPT, 6.2 Astra scored 84.5% of available credit, ahead of GPT, 6 Astra at 83.3%, GPT, 6.1 Sol at 82.0% and GPT, 5.6 Sol at 58.2%. Across 11 GPT models, OpenAI reported scores from 35.02% to 83.33% at xhigh; the results are company, reported, not an independent comparison with rival systems. The benchmark matters to drug developers because it probes decisions that can guide which compounds scientists test, but strong overall scores do not mean reliable predictions: models earned only about half the credit on compound selection and numerical, property prediction. OpenAI says MedChemBench reflects real medicinal, chemistry workflows, from structure reconstruction to lead optimization. That's all for today's AI news briefing. Have a great day!