68 articles
IBM and Hugging Face release Granite 4.0 3B Vision, a compact 3-billion-parameter multimodal model optimized for enterprise document analysis. The model combines vision and language capabilities to process images and text, designed for efficient deployment on resource-constrained hardware.
OpenAI raises $122 billion to accelerate frontier AI development, expand compute capacity, and meet growing demand for ChatGPT, Codex, and enterprise AI solutions.
Hugging Face trained language models on mRNA sequences across 25 species for $165. Models capture biological patterns and enable secondary structure and gene expression predictions without manual annotation.
Hugging Face releases TRL v1.0, a post-training library for language model fine-tuning. Version 1.0 marks API stability and includes support for DPO, PPO, and optimizations for models like Llama and Mistral.
Review of new models and organizations: Nemotron Super (NVIDIA), Sarvam (Indian models), Cohere Transcribe (transcription). Overview of latest open-source releases and emerging initiatives.
OpenAI and Gates Foundation run a workshop in Asia to help disaster response teams operationalize AI tools. Initiative focuses on practical deployment of AI for improving emergency coordination and intervention efficiency across the region.
Google DeepMind reimagines the mouse pointer as a context-aware AI assistant. The tool embeds AI collaboration directly in Chrome, removing friction from traditional prompting workflows.
STADLER, a 230-year-old rail company, deploys ChatGPT to transform knowledge work across 650 employees. OpenAI reports productivity gains and time savings without specific metrics. Standard customer case study lacking measurable benchmarks or detailed results.
Hugging Face releases OpenClaw, an open-source tool to unlock robotic manipulation capabilities. The project aims to democratize access to gripper and robotic arm control technologies through a collaborative platform.
Google DeepMind releases Gemini 3.1 Flash Live, an improved voice model with better precision and lower latency for more fluid, natural and precise voice interactions.
Google DeepMind researches AI's harmful manipulation risks across finance and health sectors, developing new safety measures to protect users.
Google DeepMind launches Lyria 3 Pro, capable of generating longer tracks with structural awareness. The tool integrates into more Google products.
OpenAI releases Model Spec, a public framework defining expected behavior for its AI models. The document balances safety, user freedom, and accountability as AI systems advance.
OpenAI launches a Safety Bug Bounty program to identify AI abuse and security risks including agentic vulnerabilities, prompt injection, and data exfiltration. Researchers receive rewards for reporting discovered flaws.
OpenAI releases prompt-based teen safety policies for developers using gpt-oss-safeguard to moderate age-specific risks in AI systems.
ChatGPT adds Agentic Commerce Protocol for product discovery with visual comparisons and merchant integration. Enables immersive shopping experience directly in the chat interface.
OpenAI Foundation commits at least $1 billion to disease research, economic opportunity, AI resilience, and community programs. Philanthropic initiative with no operational details or timeline disclosed.
Hugging Face introduces EVA, a new framework for evaluating voice agents. The framework establishes standardized metrics to measure performance, robustness, and utility of voice dialogue systems in production.
OpenAI launches Sora 2 and its app with built-in safety protections to address risks from an advanced video model and social creation platform. The approach relies on concrete moderation and control mechanisms.
Nathan Lambert argues self-improvement in AI models is real but doesn't trigger fast takeoff. Improvement is gradual with diminishing returns, contradicting exponential growth scenarios.
Hugging Face releases a guide to build a domain-specific embedding model in under a day. The method uses efficient fine-tuning techniques on domain-specific data to create optimized embeddings without expensive infrastructure.
Île-de-France Region launches a call for proposals to design, build and operate an AI-dedicated Third Place in Paris. Public-Private Partnership (PPP) scheme aimed at creating an innovation and collaboration space.
OpenAI details its chain-of-thought monitoring approach for internal coding agents to detect misalignment. Analysis of real-world deployments aims to identify risks and strengthen AI safety safeguards.
OpenAI acquires Astral to accelerate Codex growth and power next-generation Python developer tools. The acquisition aims to integrate Astral's capabilities into OpenAI's developer tooling. No financial terms or timeline disclosed.
Hugging Face releases its biannual report on the open source ecosystem. The report covers trends in models, datasets, and tools available on the platform in spring 2026, with metrics on growth and adoption.
Google DeepMind introduces a framework to measure progress toward AGI and launches a Kaggle hackathon to build relevant evaluations.
Hugging Face releases Holotron-12B, a 12-billion-parameter AI agent designed for high-throughput computer automation tasks. The model combines vision and action capabilities to interact with user interfaces.
OpenAI releases GPT-5.4 mini and nano, smaller variants of GPT-5.4 optimized for coding, tool use, multimodal reasoning, and high-volume API and sub-agent workloads. Targets cost-efficient inference at scale.
OpenAI Japan introduces Japan Teen Safety Blueprint featuring age verification, parental controls, and well-being safeguards for teen users of generative AI. Regional safety initiative with no technical specifications or rollout timeline disclosed.
OpenAI reports Americans send ~3 million daily messages to ChatGPT asking about compensation and earnings. This massive usage helps close the wage information gap in the labor market.
Nathan Lambert examines the industrialization of open-source language models: market dynamics, capability progression, adaptation strategies, and growing confusion among players facing sector consolidation.
Codex Security (OpenAI) replaces traditional SAST with AI-driven constraint reasoning and validation to detect real vulnerabilities with fewer false positives, avoiding standard SAST reports.
OpenAI outlines ChatGPT's defenses against prompt injection and social engineering in AI agent workflows. Approach: constraining risky actions and protecting sensitive data at the agent level.
OpenAI built an agent runtime using the Responses API, shell tool, and hosted containers to run secure, scalable agents with files, tools, and state. Full architecture for transforming models into autonomous agents.
Wayfair deploys OpenAI models to automate support ticket triage and enhance product attributes across millions of catalog items. Improves data accuracy and accelerates customer support response times at scale.
Rakuten fixes code issues twice as fast using OpenAI's Codex model. The e-commerce company reports measurable productivity gains in developer bug-fixing workflows. Real-world deployment of generative AI for code assistance at enterprise scale.
OpenAI introduces IH-Challenge, a training method that improves instruction hierarchy in frontier LLMs. It strengthens prioritization of trusted instructions, safety steerability, and resistance to prompt injection attacks.
ChatGPT adds interactive visual explanations for math and science, enabling students to explore formulas, variables, and concepts in real time.
Hugging Face analyzes 16 open-source reinforcement learning libraries to identify best practices in token management and data flow optimization. The study documents implementation patterns, scalability challenges, and recommendations for maintaining RL system efficiency in production.
Hugging Face introduces Storage Buckets on its Hub, enabling users to store and manage large files directly on the platform. This feature simplifies sharing of datasets, models, and artifacts without relying on external services.
Google DeepMind reflects on ten years of AlphaGo's impact, from its Go victory to applications in biology and beyond. The article explores how AlphaGo catalyzes scientific discovery and paves a path toward AGI.
OpenAI acquires Promptfoo, an AI security platform for identifying and remediating vulnerabilities in AI systems during development.
LeRobot v0.5.0 scales across all dimensions: larger models, expanded datasets, multi-robot support, and training optimizations. The open-source robotics platform reaches production maturity with standardized benchmarks and modern architecture integration.
Hugging Face introduces Ulysses, a sequence parallelism technique for training models on million-token contexts. The method distributes attention computations across multiple GPUs without reducing batch size, improving memory efficiency and training speed.
OpenAI releases Codex Security in research preview: an AI application security agent that analyzes project context to detect, validate, and patch complex vulnerabilities with higher confidence and reduced false positives.
Descript leverages OpenAI's reasoning models to automate multilingual video dubbing at scale while preserving timing and meaning of original content.
Balyasny Asset Management built an AI research engine combining rigorous model evaluation, full OpenAI platform integration, and agent workflows to reinvent investment research processes. The system leverages multi-agent architectures for financial analysis.
Allenai releases Olmo Hybrid, exploring novel LLM architectures. The article covers advances in open-source post-training tools and future directions in model design.
Hugging Face releases a comprehensive guide for deploying robotics AI models on embedded platforms. The project covers dataset recording, Vision Language Action (VLA) fine-tuning, and on-device optimizations to reduce latency and memory consumption.
OpenAI introduces CoT-Control, a tool measuring reasoning models' ability to control their chains of thought. Results show these models struggle to direct their internal processes, reinforcing monitorability as a critical AI safety safeguard.
OpenAI releases GPT-5.4, its most capable and efficient frontier model for professional work, with state-of-the-art coding, computer use, tool search, and 1M-token context window.
OpenAI releases System Card for GPT-5.4 Thinking, documenting advanced reasoning capabilities. Technical documentation covers benchmarks, limitations, and safety considerations for extended thinking system.
OpenAI releases tools, certifications, and measurement resources to help schools and universities close AI capability gaps and expand educational opportunity. Initiative focused on equitable access and AI capability training.
OpenAI launches ChatGPT for Excel with new financial data integrations powered by GPT-5.4 to accelerate modeling, research, and analysis in regulated environments.
VfL Wolfsburg (Bundesliga) deploys ChatGPT club-wide to boost efficiency, creativity, and knowledge-sharing. The strategy prioritizes training all staff rather than isolated pilots, while maintaining the club's football identity.
OpenAI launches an Adoption news channel offering practical insights and frameworks to help businesses convert AI progress into competitive advantage. No new models or technical details announced.
OpenAI outlines five AI value models for business leaders: workforce fluency, process reinvention, and building durable competitive advantage. No technical details or benchmarks provided in excerpt.
Hugging Face introduces Modular Diffusers, a modular architecture for diffusion pipelines. Reusable components enable combining schedulers, denoisers, and other building blocks to create custom image generation models without rewriting code.
Preprint extends single-minus amplitudes to gravitons, with GPT-5.2 Pro assisting in deriving and verifying nonzero graviton tree amplitudes in quantum gravity.
Axios uses AI to support local reporters, streamline newsroom workflows, and scale high-impact local journalism, per COO Allison Murphy. No specific models or technical details disclosed in excerpt.
OpenAI releases the Learning Outcomes Measurement Suite to assess AI's impact on student learning across diverse educational environments over time. The tool measures pedagogical effects longitudinally. Initiative focused on empirical evaluation of AI's educational benefits.
Hugging Face demonstrates training a full text-to-image model in 24 hours. PRX Part 3 details the infrastructure, optimizations, and techniques used to drastically reduce training time without sacrificing quality.
Google DeepMind releases Gemini 3.1 Flash-Lite, the fastest and most cost-efficient model in the Gemini 3 series. Optimized for large-scale inference with reduced latency and minimal costs.
Qwen 3.5, GLM 5, and MiniMax 2.5: Chinese labs release latest frontier models. Interconnects covers recent advances from major Chinese AI players.
OpenAI releases System Card for GPT-5.3 Instant, documenting model capabilities, limitations, and safety measures. Official technical documentation with no release date or performance metrics disclosed.
OpenAI releases GPT-5.3 Instant, a model optimized for everyday conversations with reduced latency and improved fluidity. Positioned as a faster, lighter alternative to previous versions for routine use.
The France Digitale/EY Barometer 2024 demonstrates the resilience of France's startup ecosystem despite economic and geopolitical instability. The annual study documents continued sector growth amid macroeconomic challenges.
Anthropic releases Claude Sonnet 4.5, its new model positioned as a leader in programming. The model improves code generation and reasoning capabilities compared to previous versions.