Here are today's news for 2nd of November about OpenAI, Anthropic, Meta and Google DeepMind. Let's dive into the details. OpenAI announced GPT, 6.1 Terra on October 26, 2026, for visual reasoning, computer use and writing. The model is rolling out in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, and is available to developers through the API as `gpt, 6.1, terra`; it is not offered in standard ChatGPT conversations. API pricing is $1 per million input tokens, 5 cents per million cached input tokens and $5 per million output tokens, 50% below GPT, 5.6 Terra and GPT, 6.1 Sol. OpenAI says Terra improves on its predecessor in computer, based work: in Agents’ Last Exam, it scored 15 points higher than GPT, 5.6 Terra at medium inference, at roughly one, third the task cost. It also exceeded the previous Terra’s best OSWorld 2.1 computer, use score at medium inference. The launch extends OpenAI’s GPT, 6 lineup toward lower, cost, high, volume automation, including workflows such as extracting invoice details and entering them into accounting software, while reserving Sol and Astra for more demanding tasks. On Oct. 26, 2026, Anthropic filed a public Form S, 1 with the SEC for an initial offering of Class A shares, proposing a Nasdaq Global Select Market listing under “ANT.” Neither the share count nor price range is set, and sales cannot begin until the registration statement takes effect. The filing follows a confidential submission in June and gives investors a first detailed view of the Claude maker’s finances: revenue reached $16.3 billion in the first half of 2026, while net losses were $80.7 billion. The gap highlights the cost and risk behind a generative, AI boom increasingly fueled by paid business use, even as rivals race to scale. Anthropic’s filing also describes more than $500 billion in computing commitments, much of it difficult to cancel, underscoring how access to chips and data, center capacity can shape competition. The IPO could test public investors’ appetite for AI companies whose growth is exceptional but spending and losses are immense; its timing and valuation remain unsettled. On Oct. 26, Meta said it would seek up to $95 billion in equity to fund AI infrastructure and “personal superintelligence.” Its initial roughly $35 billion offering comprised $22.5 billion of Class A common stock and $12.5 billion of mandatory, convertible preferred shares; a proposed $60 billion at, the, market sale of non, voting Class C shares could begin after they start trading in Q1 2027. The common shares were later priced at $625 each, and Meta increased that portion to $25 billion, bringing the announced total to $97.5 billion. Nvidia committed $3 billion, while chipmakers Nvidia, AMD and Broadcom showed interest in up to $5 billion combined. The scale highlights how AI infrastructure financing is becoming a competitive battleground, and could dilute existing shareholders. Meta forecasts $135 billion to $145 billion in 2026 capital expenditure, building on its expansion of data centers and computing capacity. Google DeepMind announced Gemini 4 Carbon on October 28, 2026, calling it the company’s strongest model and highlighting long, horizon work, science and cybersecurity. Google says it will initially provide access to a subset of cyber defenders, then expand to trusted partners through its Fairwind program; it has not announced a date for general availability. The staged rollout reflects the model’s dual, use capabilities: defenders may use it to find and fix software flaws, while broader access is being controlled. Google reports 96.2% on GraphWalks’ 256,000, to, 1, million, token long, context test, versus 84.2% for Gemini 4 Argon and 71.8% for GPT, 6 Astra. Carbon scored 53.9% on AutomationBench, 80.0% on RiemannBench and 64.8% on Terminal, Bench, Science. In security evaluations, it reached 88.2% on SEC, bench Pro for vulnerability discovery and 71.0% on CWE, bench for patching. These are company, reported benchmarks, not independent proof of real, world performance. On October 26, OpenAI said an unnamed internal model had found an algorithm refuting the Exponential Time Hypothesis (ETH), which holds that 3, SAT cannot be solved in subexponential time. Anthropic published a similar claim hours later in “A 2^o(n) Time Algorithm for 3, SAT,” credited to Claude. Both companies say their models reached the core result on October 16 and formalized proofs in Lean, a system that checks logical steps. Anthropic reports its model used 14 million output tokens, with no human input to the discovery. The work also claims faster exact counting of satisfying assignments, with a bound of 2^O(n log log n/√log n). If independently confirmed, the result would unsettle assumptions used across complexity theory and hundreds of research papers; it does not establish practical speedups or resolve P versus NP. ETH’s original proposers and MIT’s Ryan Williams are helping organize scrutiny, but neither proof has been peer, reviewed, and researchers have yet to verify that the formal statements match the intended conjecture. That's all for today's AI news briefing. Have a great day!