Curated releases and announcements from OpenAI, Anthropic, and more.
OpenAI has introduced Advanced Account Security, an opt-in feature for ChatGPT and Codex that combines phishing-resistant sign-in, restricted account recovery, shorter sessions, and automatic training exclusion to protect users at heightened risk of digital attacks.
Anthropic has released a set of connectors that integrate Claude with major creative software tools-including Blender, Adobe, Autodesk, Ableton, and Splice-enabling creative professionals to use natural-language AI assistance directly within their existing workflows.
OpenAI and AWS have deepened their strategic partnership by bringing OpenAI's models (including GPT-5.5), Codex, and Amazon Bedrock Managed Agents to AWS, enabling enterprises to build with OpenAI's AI capabilities within their existing AWS infrastructure, security, and compliance workflows.
Anthropic has hired Theo Hourmouzis, a veteran technology executive with over 20 years of Asia Pacific experience, as General Manager of Australia and New Zealand, while officially launching the company's Sydney office and announcing new partnerships with Canva, Xero, and YMCA South Australia.
NEC Corporation is adopting Claude across approximately 30,000 employees globally and partnering with Anthropic as its first Japan-based global partner to co-develop secure, industry-specific AI solutions for sectors including finance, manufacturing, and government.
Anthropic details the measures it is taking to ensure Claude provides accurate, impartial, and reliable election-related information ahead of the 2026 US midterms and other global elections, including bias evaluations, policy enforcement, influence operation testing, election resource banners, and web search integration.
Google's Gemini can help with seasonal cleaning and everyday organization - from custom cleaning plans and fridge recipe ideas to repair troubleshooting, errands planning, room redesign previews, plant care and inbox decluttering.
OpenAI has released GPT-5.5, the company's most capable model yet, which significantly advances agentic coding, knowledge work, and scientific research while matching GPT-5.4's latency and using fewer tokens. The model is rolling out across ChatGPT and Codex with enhanced cybersecurity safeguards and will reach the API soon.
OpenAI has released workspace agents in ChatGPT, allowing teams to build Codex-powered shared agents that handle complex, multi-step workflows across tools like Slack and CRMs-all within organizational permissions and controls.
OpenAI has released ChatGPT for Clinicians, a free specialized version of ChatGPT for verified U.S. physicians, NPs, PAs, and pharmacists, designed to support clinical documentation, medical research, and care consultation. The launch is accompanied by HealthBench Professional, an open benchmark for evaluating AI performance on real clinician tasks.
OpenAI has released ChatGPT Images 2.0, a major upgrade to its image generation capabilities that delivers greater precision, enhanced realism across diverse visual styles, improved multilingual text rendering, flexible aspect ratios, and the ability to act as an end-to-end visual thought partner.
Anthropic has entered into an expanded agreement with Amazon to secure up to 5 gigawatts of compute capacity for Claude, backed by a more than $100 billion commitment to AWS technologies over ten years. Amazon is also investing an additional $5 billion in Anthropic, with up to $20 billion more planned for the future.
Anthropic has launched Claude Design, an Anthropic Labs product powered by Claude Opus 4.7 that lets users collaborate with Claude to create polished visual work including designs, prototypes, presentations, and marketing collateral. It is available in research preview for Claude Pro, Max, Team, and Enterprise subscribers.
Anthropic has released Claude Opus 4.7, a significant upgrade over Opus 4.6 that delivers major improvements in advanced software engineering, instruction following, high-resolution vision, and long-running agentic tasks, along with new cybersecurity safeguards and developer features like the xhigh effort level and ultrareview command.
OpenAI has released a major Codex update that expands it well beyond code generation-adding computer use, image generation, memory, scheduling, 90+ plugins, and deeper developer workflow support to serve as a more complete partner across the entire software development lifecycle.
OpenAI has released a major update to its Agents SDK, adding a model-native harness with configurable memory, filesystem tools, and standardized integrations, along with native sandbox execution that lets agents safely inspect files, run commands, and handle long-horizon tasks in controlled environments.
Anthropic's Long-Term Benefit Trust has appointed Vas Narasimhan, CEO of Novartis and a physician-scientist, to Anthropic's Board of Directors, making Trust-appointed directors a majority on the board. Narasimhan brings extensive experience in stewarding breakthrough science responsibly across healthcare and life sciences.
As moving season picks up, Google's Gemini can help with visualizing renovations, summarizing dense documents, refining listing searches, finding emails, researching neighborhoods and modeling mortgage scenarios.
Anthropic has entered into an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity starting in 2027, representing the company's largest compute commitment yet as run-rate revenue surpasses $30 billion and the number of million-dollar business customers doubles to over 1,000.
OpenAI now allows ChatGPT Business and Enterprise teams to add Codex-only seats with pay-as-you-go pricing, while also lowering the annual ChatGPT Business seat price from $25 to $20.
Anthropic has signed a Memorandum of Understanding with the Australian government to collaborate on AI safety research, share insights on frontier model capabilities, and support Australia's National AI Plan. The agreement also includes AUD$3 million in research partnerships with Australian institutions focused on genomics, rare disease diagnosis, and computing education.
OpenAI is rolling out richer, more visually immersive shopping experiences in ChatGPT, powered by an expansion of the Agentic Commerce Protocol (ACP) that brings product discovery, side-by-side comparisons, and up-to-date merchant catalog data directly into conversations.
OpenAI has released GPT-5.4 mini and nano, fast and efficient small models optimized for coding, subagents, and computer use that bring much of GPT-5.4's capability at significantly lower latency and cost.
OpenAI explains why Codex Security was intentionally designed to analyze repositories from the code and system intent rather than starting from SAST reports, since the most critical vulnerabilities involve defenses that appear correct but fail to enforce the properties the system depends on.
Anthropic launched the Claude Partner Network with an initial $100 million commitment for 2026, providing training, technical certifications, dedicated engineering support, and joint market development to help partner organizations accelerate enterprise adoption of Claude.
Anthropic has established The Anthropic Institute, an interdisciplinary initiative led by co-founder Jack Clark, aimed at studying and communicating the societal, economic, and security challenges posed by increasingly powerful AI systems. The Institute combines Anthropic's Frontier Red Team, Societal Impacts, and Economic Research groups while also expanding the company's Public Policy organization.
OpenAI has launched interactive visual explanations in ChatGPT covering more than 70 math and science concepts, allowing learners to manipulate variables and see real-time effects on formulas and graphs across all plans.
Anthropic is opening a new office in Sydney, marking its fourth Asia-Pacific location alongside Tokyo, Bengaluru, and Seoul, to meet growing demand from Australian and New Zealand businesses and deepen partnerships across the region.
Anthropic collaborated with Mozilla to use Claude Opus 4.6 for discovering 22 new vulnerabilities in Firefox over two weeks, 14 of which were rated high-severity, demonstrating that AI models can now serve as world-class security researchers while also raising concerns about emerging AI-enabled exploit development.
OpenAI has released Codex Security in research preview, an application security agent that uses frontier AI models and deep project context to find complex vulnerabilities with high confidence, propose actionable fixes, and minimize false positives - now available to ChatGPT Enterprise, Business, and Edu customers.
OpenAI has released GPT-5.4, its most capable and efficient frontier model designed for professional work, featuring native computer-use capabilities, 1M token context support, improved tool search, and significantly better performance on knowledge work, coding, and agentic workflows across ChatGPT, Codex, and the API.
OpenAI has released ChatGPT for Excel in beta, an add-in powered by GPT‑5.4 that lets users build, update, and analyze spreadsheet models directly within workbooks, alongside new financial data integrations from providers like FactSet, Moody's, and S&P Global.
Anthropic's CEO Dario Amodei responded to the Department of War's supply chain risk designation, stating the company plans to challenge it in court while emphasizing that the vast majority of customers are unaffected and that Anthropic remains committed to supporting national security.
OpenAI has released GPT-5.3 Instant, an update to ChatGPT's most-used model that improves conversational tone, reduces unnecessary refusals and hallucinations, delivers better web-sourced answers, and offers stronger writing capabilities.
Anthropic responded to Secretary of War Pete Hegseth's announcement directing the Department of War to designate the company as a supply chain risk, following a negotiation impasse over Anthropic's refusal to allow its AI model Claude to be used for mass domestic surveillance and fully autonomous weapons. The company stated it would challenge the designation in court and assured customers that their access to Claude would remain unaffected outside of Department of War contracts.
Anthropic's CEO Dario Amodei has outlined the company's extensive AI deployment across national security agencies while affirming that Anthropic will not agree to the Department of War's demands to remove safeguards against mass domestic surveillance and fully autonomous weapons, despite threats of removal from government systems.
Anthropic has acquired Vercept, a startup specializing in AI perception and interaction, to bolster Claude's computer use capabilities as the model approaches human-level performance on complex software tasks.
Anthropic has released the third version of its Responsible Scaling Policy, restructuring the framework to separate unilateral company commitments from industry-wide recommendations, while introducing Frontier Safety Roadmaps, periodic Risk Reports, and external review mechanisms to improve transparency and accountability.
Quick guidance for using Lyria 3 in the Gemini app to create 30-second songs from text, images or videos, with prompt examples and sharing tips.
Anthropic revealed that DeepSeek, Moonshot AI, and MiniMax conducted industrial-scale distillation campaigns using over 24,000 fraudulent accounts to illicitly extract Claude's capabilities, generating more than 16 million exchanges targeting agentic reasoning, coding, and tool use. Anthropic is responding with enhanced detection systems, intelligence sharing, and stricter access controls, while calling for coordinated industry and policy action.
Anthropic has released Claude Code Security in a limited research preview, offering AI-powered codebase scanning that reasons about code like a human security researcher to find complex vulnerabilities traditional tools miss. Enterprise, Team, and open-source maintainers can apply for early access to review findings and approve suggested patches through a dedicated dashboard.
Anthropic has partnered with Infosys to develop and deploy AI agents specifically designed for telecommunications, financial services, manufacturing, and software development industries. The collaboration combines Claude models with Infosys Topaz to deliver enterprise AI solutions that meet the stringent governance and compliance requirements of regulated sectors.
Rwanda and Anthropic have formalized a three-year partnership through a Memorandum of Understanding to deploy AI across healthcare, education, and government services. This collaboration, Anthropic's first multi-sector government MOU in Africa, includes supporting Rwanda's health initiatives, providing Claude access to government developers, and expanding educational programs across eight African countries.
Anthropic releases Claude Sonnet 4.6, featuring major improvements in coding, computer use, and long-context reasoning with a 1M token context window. The model shows strong developer preference over previous versions and even outperforms Opus 4.5 in many scenarios while maintaining the same pricing.
Anthropic has opened its Bengaluru office and unveiled extensive partnerships across Indian enterprises, education, and agriculture sectors. The company is enhancing Claude's capabilities in ten Indian languages while collaborating with organizations like Air India, CRED, Pratham, and EkStep Foundation to bring AI benefits to millions across India.
Anthropic has partnered with CodePath to integrate Claude into computer science curricula for over 20,000 students at community colleges, state schools, and HBCUs. The collaboration aims to democratize access to AI development tools and reshape programming education for students from underserved communities.
Anthropic has appointed veteran executive Chris Liddell, former CFO of Microsoft and General Motors and former Deputy White House Chief of Staff, to its Board of Directors. His extensive experience in technology leadership, public service, and governance will help guide Anthropic's mission to develop AI that is both capable and responsible.
OpenAI has introduced two major security features for ChatGPT to combat prompt injection attacks: Lockdown Mode for high-risk users requiring enhanced protection, and "Elevated Risk" labels to help users make informed decisions about potentially vulnerable features. These protections build upon OpenAI's existing security infrastructure and will initially be available to enterprise customers before expanding to consumers.
Anthropic has announced a $20 million donation to Public First Action, a bipartisan organization advocating for responsible AI governance policies. The contribution supports efforts to establish transparency requirements, maintain American AI leadership, and implement safeguards against AI-related risks while enabling the technology's benefits.
OpenAI launches GPT-5.3-Codex-Spark, an ultra-fast model delivering over 1,000 tokens per second for real-time coding through partnership with Cerebras. The model enables near-instant collaboration and rapid iteration while maintaining strong coding capabilities.
Anthropic has raised $30 billion in Series G funding at a $380 billion valuation, led by GIC and Coatue, with the company reporting $14 billion in run-rate revenue and over 10x annual growth. The funding will support frontier research, product development, and infrastructure expansion as Claude becomes increasingly critical for enterprise AI and coding applications.
Anthropic announces comprehensive commitments to cover electricity price increases resulting from its data center operations, including full funding of grid infrastructure costs and investments in new power generation. The initiative aims to prevent American consumers from bearing the financial burden of AI infrastructure expansion while supporting federal policies for systemic energy improvements.
OpenAI has launched advertising tests in ChatGPT for U.S. users on Free and Go tiers to fund infrastructure while maintaining user privacy and answer independence. The ads system includes user controls, privacy protections, and clear separation from organic responses, with revenue supporting broader access to AI capabilities.
OpenAI unveils Frontier, a comprehensive enterprise platform that enables organizations to build, deploy, and manage AI agents with shared context, learning capabilities, and proper governance. The platform helps enterprises bridge the gap between AI potential and practical deployment, with early adopters including HP, Oracle, and State Farm reporting significant operational improvements.
Anthropic has released Claude Opus 4.6, featuring state-of-the-art performance in agentic coding, extended context handling up to 1M tokens, and dramatically improved capabilities for autonomous task completion. The model achieves industry-leading benchmarks while maintaining robust safety standards and introduces new developer features including adaptive thinking and context compaction.
OpenAI has released GPT-5.3-Codex, a groundbreaking AI model that extends beyond code generation to handle virtually any professional computer task. The model, which helped develop itself, achieves state-of-the-art performance across coding, web development, and knowledge work benchmarks while introducing real-time collaborative features.
Anthropic announces that Claude will remain permanently free from advertisements, prioritizing user trust and genuine assistance over advertising revenue. The company believes ads would compromise Claude's role as a dedicated workspace for serious thinking and professional tasks.
Apple's Xcode 26.3 now integrates Anthropic's Claude Agent SDK, enabling autonomous coding capabilities including visual verification, cross-project analysis, and independent task execution within the IDE. This enhancement allows developers to leverage Claude Code's full functionality for complex, extended programming tasks without leaving Xcode.
OpenAI outlines the design philosophy behind Sora's content feed, emphasizing creativity-focused ranking, user control, and community building. The platform balances safety measures with creative freedom through multi-layered content moderation and personalized recommendations.
OpenAI has released the Codex desktop app for macOS, a powerful interface that enables developers to manage multiple AI agents simultaneously, run parallel tasks, and collaborate on long-running projects. The app includes features like skills for extending functionality beyond code generation, automated background processing, and customizable interaction styles.
Anthropic announces partnerships with the Allen Institute and Howard Hughes Medical Institute to develop AI systems that accelerate biological research and scientific discovery. These collaborations will integrate Claude into experimental workflows, creating specialized AI agents and multi-agent systems to help scientists analyze complex data, generate hypotheses, and design experiments more efficiently.
OpenAI will remove GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini from ChatGPT on February 13, 2026, following user migration to newer models. The decision comes after incorporating GPT-4o's popular features into GPT-5.2, with only 0.1% of users still actively using the older model.
ServiceNow has adopted Claude as its default AI model for enterprise application development and internal operations, enabling customers to build automated workflows while reducing implementation time by 50%. The company has also deployed Claude across its 29,000-person workforce, achieving 95% reduction in sales preparation time and enhanced engineering productivity.
Anthropic has been selected by the UK government to develop an AI assistant for GOV.UK that will help citizens navigate government services, starting with employment support. The partnership includes knowledge transfer to government staff and represents part of Anthropic's growing investment in the UK's AI leadership position.
A Google AI: Release Notes episode explores how the Smokejumpers team developed Gemini’s code model and scaled its serving infrastructure, with product and research leads discussing design goals, “vibe coding,” and AI’s impact on programming.
OpenAI has released Prism, a free AI-powered workspace that integrates GPT-5.2 directly into scientific writing and collaboration workflows. The platform combines LaTeX editing, real-time teamwork, and advanced AI reasoning in a unified cloud-based environment accessible to all ChatGPT users.
Anthropic has published a comprehensive new constitutional document for Claude that explains the AI model's values, behavioral guidelines, and operational context. The document shifts from rule-based directives to narrative explanations of why Claude should behave certain ways, emphasizing safety, ethics, compliance, and helpfulness while addressing philosophical questions about AI nature and consciousness.
Anthropic has released a comprehensive update to their Responsible Scaling Policy, introducing more flexible risk assessment approaches and refined safeguards for managing catastrophic AI risks. The revision includes new capability thresholds for autonomous AI research and CBRN weapons assistance, enhanced security standards, and improved governance processes based on lessons learned from a year of implementation.
Google's Gemini offers free, full-length SAT practice exams featuring content from The Princeton Review to help students prepare.
Mariano-Florentino Cuéllar has been appointed to Anthropic's Long-Term Benefit Trust, bringing extensive experience in law, governance, and international affairs to help guide the company's public benefit mission. The Trust also announced the departure of founding members Kanika Bahl and Zachary Robinson after their significant contributions to establishing its governance role.
Anthropic collaborates with Teach For All to provide AI training and Claude access to over 100,000 educators across 63 countries. The partnership enables teachers to co-create AI tools tailored to classroom needs while serving more than 1.5 million students globally.
ServiceNow has formed a multi-year partnership with OpenAI to integrate advanced AI models including GPT-5.2 into its enterprise workflow platform, serving over 80 billion annual workflows. This collaboration enables businesses to deploy AI that understands, decides, and takes automated action within secure enterprise environments across all major business functions.
OpenAI has launched ChatGPT Go worldwide, an $8 monthly subscription offering expanded access to GPT-5.2 Instant with 10x more messages and uploads than the free tier. The service includes plans for ad testing to maintain affordability.
OpenAI announces plans to introduce advertisements in ChatGPT's free and Go tiers while launching the $8/month Go subscription globally. The company outlines five core principles for advertising implementation, emphasizing user privacy, answer independence, and maintaining user trust while expanding AI accessibility.
Anthropic has appointed Irina Ghose, former Microsoft India Managing Director with over 30 years of technology leadership experience, to lead its India operations as the company establishes its first office in Bengaluru. The appointment comes as India represents Claude.ai's second-largest global market, with Indian users primarily focused on technical and mathematical applications.
Researchers are developing custom Claude-powered systems that dramatically accelerate scientific workflows, from analyzing massive genomic datasets in minutes instead of months to navigating hundreds of biomedical tools automatically. These AI collaborations are enabling scientists to eliminate research bottlenecks and discover patterns that human experts might miss.
A late-night codename suggestion combining a Googler’s nicknames became the public moniker for Google's image-editing model; the playful name helped the model go viral and influenced product branding.
Anthropic has expanded its Labs division to develop experimental products leveraging Claude's advanced capabilities, with Mike Krieger joining Ben Mann to lead innovation efforts. This strategic move follows successful launches like Claude Code and MCP, positioning Anthropic to explore AI's frontier while scaling reliable products for enterprise and consumer markets.
Anthropic launches Claude for Healthcare with HIPAA-ready tools and expands life sciences capabilities, featuring new platform integrations for clinical trials and regulatory operations. The enhanced Claude Opus 4.5 model demonstrates significant improvements in medical benchmarks and scientific task performance.
OpenAI has introduced a comprehensive suite of AI tools for healthcare organizations, including ChatGPT for Healthcare and API services, designed to improve patient care quality while maintaining HIPAA compliance. The solutions are already being deployed at major medical institutions like Boston Children's Hospital and UCSF, helping clinicians reduce administrative burden and access evidence-based medical knowledge more efficiently.
OpenAI launches ChatGPT Health, a secure feature that integrates medical records and wellness apps to provide personalized health guidance. The dedicated space offers enhanced privacy protections and was developed in collaboration with over 260 physicians worldwide.
Google's Nano Banana Pro, powered by Gemini 3 Pro, is an upgraded image generation and editing model available across the Gemini app, AI Mode in Search, NotebookLM, Workspace (Slides and Vids), Flow, Mixboard, and developer platforms like Vertex AI.
Anthropic has released its Frontier Compliance Framework to meet California's new AI transparency law (SB 53), which mandates frontier AI developers disclose how they assess and manage catastrophic risks. The company advocates for federal legislation to establish consistent nationwide standards for AI safety and transparency.
Google's December 2025 Gemini app update introduces Gemini 3 Flash globally, more precise image editing with Nano Banana, NotebookLM integration for notebook sources, visual Deep Research reports for Ultra users, and richer local results from Google Maps.
Anthropic has announced a multi-year collaboration with the US Department of Energy through the Genesis Mission initiative to apply AI across energy, life sciences, and scientific research. The partnership will provide DOE researchers with access to Claude and specialized AI tools to maintain America's scientific leadership and advance critical research areas.
OpenAI launches GPT-5.2-Codex, featuring breakthrough capabilities in complex software development and defensive cybersecurity applications. The model achieves state-of-the-art performance on coding benchmarks while introducing enhanced safeguards and a trusted access program for qualified security professionals.
OpenAI has launched app submissions for ChatGPT, allowing developers to create and publish applications that enhance conversations with additional context and actions. The company has also introduced an app directory within ChatGPT for users to discover and connect with these applications.
Google's Gemini 3 Flash is a faster, lower-cost addition to the Gemini 3 family that delivers advanced reasoning and multimodal capabilities with low latency, now rolling out across Google apps, Search and developer platforms.
OpenAI has released ChatGPT Images powered by GPT Image 1.5, featuring 4x faster generation speeds, more precise editing capabilities, and a new dedicated creative interface. The model delivers improved consistency in preserving facial features and details across edits while offering enhanced text rendering and natural-looking outputs for both consumer and business applications.
OpenAI unveils GPT-5.2, establishing new benchmarks in professional knowledge work with 70.9% performance matching or exceeding industry professionals on real-world tasks. The model brings substantial improvements in coding, long-context understanding, vision capabilities, and tool usage while reducing hallucinations by 30% compared to previous versions.
OpenAI has released GPT-5.2 Pro and GPT-5.2 Thinking, which demonstrate significant improvements in mathematical and scientific problem-solving capabilities. The models have achieved state-of-the-art performance on graduate-level science benchmarks and successfully contributed to solving a previously open problem in statistical learning theory.
Disney and OpenAI have formed a three-year partnership enabling Sora to generate AI videos featuring over 200 Disney, Marvel, Pixar, and Star Wars characters. The agreement includes a $1 billion Disney investment in OpenAI and plans to stream fan-created content on Disney+ starting in 2026.
Anthropic launches Claude Opus 4.5, a state-of-the-art AI model excelling in coding, agents, and computer use with improved efficiency and safety. The model outperforms competitors on software engineering benchmarks while using fewer tokens, and is now available at $5/$25 per million tokens through Anthropic's API and platforms.
Google's Gemini 3 update brings stronger reasoning, model-generated interactive interfaces, a redesigned app and an experimental Gemini Agent that can perform multi-step tasks for users.
OpenAI releases GPT-5.1 Instant and GPT-5.1 Thinking models with enhanced intelligence, improved conversational abilities, and expanded personalization options for ChatGPT users. The update includes adaptive reasoning capabilities and customizable tone settings, rolling out first to paid subscribers before expanding to all users.
Anthropic launches Claude Haiku 4.5, a small model that delivers near-frontier coding performance at one-third the cost and double the speed of previous state-of-the-art models. The model provides high intelligence with exceptional speed for real-time applications while maintaining strong safety standards.
OpenAI has released Sora 2, an advanced video and audio generation model with improved physical accuracy and controllability, alongside a new social iOS app featuring the ability to insert real people into generated scenes. The platform emphasizes user wellbeing with natural language-controlled feeds and comprehensive safety measures while offering powerful creative tools for video synthesis.
Anthropic launches Claude Sonnet 4.5, their most capable model excelling in coding, reasoning, and computer use with state-of-the-art performance on software development benchmarks. The release includes major product upgrades, the Claude Agent SDK for developers, and represents Anthropic's most aligned frontier model with significant safety improvements.
Anthropic has raised $13 billion in Series F funding led by ICONIQ, achieving a $183 billion valuation. The company's annualized revenue has grown from $1 billion to over $5 billion in just eight months, serving more than 300,000 business customers.
OpenAI has released GPT-5, a unified AI system featuring state-of-the-art performance in coding, mathematics, writing, and health applications with significantly reduced hallucinations and improved safety measures. The model automatically selects between quick responses and deeper reasoning based on task complexity, while GPT-5 Pro offers extended reasoning for the most challenging problems.
The White House released a comprehensive AI strategy plan that aligns with many of Anthropic's policy recommendations, though the company emphasizes the need for stronger transparency standards and export controls. Anthropic supports the plan's focus on infrastructure development, federal adoption, and safety measures while expressing concern about recent H20 chip export reversals to China.
OpenAI has released o3 and o4-mini, their most advanced reasoning models to date, featuring multimodal capabilities, comprehensive tool integration, and significant performance improvements. The models can now reason directly with images, use multiple tools strategically, and achieve state-of-the-art results across coding, mathematics, science, and visual perception tasks while maintaining cost efficiency.
OpenAI has released GPT-4.5 as a research preview, representing their most advanced chat model with improved world knowledge, reduced hallucinations, and enhanced emotional intelligence. The model, available to Pro users and developers, demonstrates significant improvements in factual accuracy, creative tasks, and human collaboration compared to previous versions.
OpenAI has published a comprehensive system card detailing the safety measures, risk mitigations, and deployment preparations for Sora, their video generation model. The document outlines extensive testing with external red teams, multi-layered content moderation systems, and specific protections for child safety, preventing deepfakes, and managing deceptive content.
OpenAI has released a comprehensive safety evaluation of GPT-4o, detailing risk assessments across audio, text, and vision capabilities with particular focus on novel audio-related risks. The model received medium risk rating for persuasion capabilities and low risk ratings for cybersecurity, biological threats, and autonomy according to OpenAI's Preparedness Framework.
OpenAI has introduced GPT-4o mini, a highly cost-effective small model that delivers superior performance at 15 cents per million input tokens. The model achieves 82% on MMLU benchmarks while being 60% cheaper than GPT-3.5 Turbo, making advanced AI capabilities significantly more accessible.