OpenAI Analysis of Model Behavior: The 'Goblin' Lexical Tic
OpenAI identified that a reward signal for the 'Nerdy' personality feature caused models from GPT-5.1 to GPT-5.5 to develop an unintended habit of using goblin and gremlin metaphors, which then generalized across other model behaviors.
IBM Granite 4.1 LLMs release notes / technical overview
IBM has released Granite 4.1, a family of dense, decoder-only LLMs (3B, 8B, and 30B) that achieve high performance through rigorous data curation and a multi-stage reinforcement learning pipeline.
OpenAI Stargate Compute Infrastructure Update
OpenAI has surpassed its initial 10GW AI infrastructure goal for the United States and trained GPT-5.5 at its flagship Stargate site in Abilene, Texas.
OpenAI Cybersecurity Action Plan for the Intelligence Age
OpenAI has introduced a five-pillar Action Plan to democratize AI-powered cyber defense and strengthen resilience against AI-driven threats.
DeepInfra Integration with Hugging Face Inference Providers
Hugging Face has added DeepInfra as a supported Inference Provider, enabling serverless access to over 100 models, including DeepSeek V4 and Kimi-K2.6, directly through the Hub and client SDKs.
NVIDIA Nemotron 3 Nano Omni release notes / what's new
NVIDIA has released Nemotron 3 Nano Omni, an omni-modal model capable of long-context reasoning across text, images, video, and audio, delivering best-in-class accuracy on document intelligence and video understanding benchmarks.
FlashQLA: CP-/Bwd-Friendly Fused Linear Attention Kernels for GDN
Qwen has open-sourced FlashQLA, a high-performance linear attention kernel library built on TileLang that achieves 2-3x forward and 2x backward speedups for Gated Delta Network (GDN) layers on NVIDIA Hopper GPUs.
NVIDIA Nemotron 3 Nano Omni Support in vLLM
vLLM now supports NVIDIA Nemotron 3 Nano Omni, a highly efficient 30B MoE multimodal model that unifies vision, audio, and language reasoning in a single loop to power agentic AI.
OpenAI Community Safety and Violence Mitigation Framework
OpenAI has detailed its multi-layered approach to preventing the use of ChatGPT for planning or executing violence, combining model training, automated detection, and human review.
OpenAI GPT-5.5, Codex, and Managed Agents on AWS release notes
On April 28 2026, OpenAI announced that its frontier models including GPT-5.5, Codex coding agent, and Amazon Bedrock Managed Agents powered by OpenAI are now available in limited preview on AWS, enabling enterprises to use these capabilities within their existing AWS security, compliance, and procurement workflows.
OpenAI achieves FedRAMP Moderate authorization
OpenAI has achieved FedRAMP 20x Moderate authorization for ChatGPT Enterprise and its API Platform, enabling U.S. government agencies to access frontier AI models including GPT-5.5 with federal security and privacy standards.
Google DeepMind Partnership with the Republic of Korea
Google DeepMind has partnered with South Korea's Ministry of Science and ICT to establish an AI Campus in Seoul and deploy frontier AI models to accelerate scientific research in life sciences, energy, and climate.
The Next Phase of the Microsoft OpenAI Partnership
OpenAI and Microsoft have amended their partnership agreement to provide greater flexibility, allowing OpenAI to serve products across any cloud provider while maintaining Microsoft as its primary cloud partner.
Choco automates food distribution with AI agents
Choco announced AI-powered OrderAgent and VoiceAgent built with OpenAI APIs to automate food distribution order processing.
OpenAI Privacy Filter: Building Scalable PII Detection Web Apps
OpenAI has released Privacy Filter, an open-source 1.5B-parameter PII detector capable of labeling eight categories of sensitive data across a 128k context window.
Symphony: Open‑Source Spec for Codex Orchestration
OpenAI released Symphony, an open‑source specification that turns issue trackers like Linear into always‑on orchestrator for Codex coding agents, enabling teams to automate routine implementation work and increase PR throughput.
OpenAI Operating Principles for AGI Development
OpenAI has outlined five core principles—Democratization, Empowerment, Resilience, Universal Prosperity, and Adaptability—to ensure that artificial general intelligence (AGI) benefits all of humanity.
OpenAI AI Jobs Transition Framework
OpenAI has introduced the AI Jobs Transition Framework to analyze how AI affects employment across 921 occupations, categorizing jobs into four paths based on automation risk, reorganization, growth potential, and stability.
vLLM DeepSeek V4 Support: Efficient Long-context Attention
vLLM now supports DeepSeek V4-Pro and V4-Flash, implementing a new attention mechanism that enables context lengths up to one million tokens with significant KV cache memory savings.
DeepSeek-V4 Release Notes: Efficient 1M-Token Context for AI Agents
DeepSeek-V4 introduces a 1M-token context window powered by a hybrid CSA/HCA attention mechanism, specifically optimized for long-running agentic workloads and tool-use trajectories.
OpenAI GPT-5.5 Release Notes
OpenAI has released GPT-5.5 and GPT-5.5 Pro, introducing significant advancements in agentic coding, computer use, and scientific research while maintaining the latency of GPT-5.4.
OpenAI GPT-5.5 System Card
OpenAI has released GPT-5.5, a model optimized for complex real-world tasks and tool use, featuring enhanced task understanding and a more robust safety framework.
How to use ChatGPT Work for everyday tasks
OpenAI’s April 23, 2026 Academy post explains how ChatGPT Work turns existing work materials—such as calendars, emails, documents, and spreadsheets—into first‑draft briefs, summaries, decks, workbooks, plans, and process documents that teams can review and edit.
Using Transformers.js in a Chrome Extension
Hugging Face provides a technical guide on integrating Transformers.js into a Chrome Extension using Manifest V3, featuring a background-hosted model architecture powered by Gemma 4 E2B.
ChatGPT for Clinicians Release
OpenAI has launched ChatGPT for Clinicians, a free version of ChatGPT for verified U.S. healthcare providers designed to automate documentation, medical research, and clinical workflows.
Google DeepMind Decoupled DiLoCo
Google DeepMind introduced Decoupled DiLoCo, a distributed training architecture that enables resilient, asynchronous LLM training across distant data centers using low bandwidth and mixed hardware generations.
Introducing workspace agents in ChatGPT – OpenAI research preview
OpenAI introduced workspace agents in ChatGPT, enabling teams to create shared, cloud‑run AI agents that automate multi‑step workflows across tools while respecting organizational controls.
Workspace agents in ChatGPT: Overview and How to Build Them
OpenAI introduced workspace agents in ChatGPT to automate repeatable, structured workflows by combining triggers, processes, and tool integrations, enabling teams to share consistent AI‑driven tasks.
OpenAI Responses API WebSocket Mode Release
OpenAI has introduced WebSocket support to the Responses API, reducing agentic workflow latency by up to 40% by eliminating redundant API overhead and enabling inference speeds of up to 4,000 tokens per second.
Qwen3.6-27B release notes / what's new
Qwen has released Qwen3.6-27B, a dense 27-billion-parameter multimodal model that outperforms the larger Qwen3.5-397B-A17B on all major agentic coding benchmarks.
OpenAI Privacy Filter Release
OpenAI has released Privacy Filter, an open-weight, 1.5B parameter model designed for high-throughput, context-aware detection and redaction of personally identifiable information (PII) in unstructured text.
vLLM FP8 KV-Cache and Attention Quantization Update
vLLM has optimized FP8 KV-cache and attention quantization to reduce decode latency and memory usage while maintaining near-baseline accuracy across Hopper and Blackwell architectures.
Google DeepMind partners with Accenture, Bain & Company, BCG, Deloitte, and McKinsey to accelerate AI transformation
Google DeepMind announced a partnership with Accenture, Bain & Company, BCG, Deloitte, and McKinsey to accelerate enterprise AI transformation by providing early access to frontier models like Gemini, co‑developing industry‑specific AI solutions, and connecting leadership with customer CEOs.
ChatGPT Images 2.0 Release
OpenAI has released ChatGPT Images 2.0, a significant update to its image generation capabilities featuring improved precision, multilingual text rendering, and enhanced stylistic realism.
QIMMA: A Quality-First Arabic LLM Leaderboard
Hugging Face and partners introduced QIMMA, a new Arabic LLM leaderboard that implements a rigorous quality validation pipeline to ensure benchmarks reflect genuine language capability.
vLLM v0.20.0 adds disaggregated serving for hybrid SSM models
vLLM v0.20.0 introduces disaggregated prefill/decode serving for hybrid SSM-FA models, enabling efficient KV transfer via dual descriptor views and a 3-descriptor conv state transfer.
Hugging Face: AI and the Future of Cybersecurity
Hugging Face argues that open-source AI models and tooling are critical for cybersecurity defense to counter the risks posed by autonomous vulnerability-finding systems like Mythos.
Scaling Codex to Enterprises Worldwide – OpenAI Announcement
OpenAI announced that Codex usage grew from over 3 million to over 4 million developers weekly and introduced Codex Labs plus a global systems‑integrator partner program to help enterprises adopt Codex across the software development lifecycle and beyond.
GRASP: Gradient-based Planning for World Models at Longer Horizons
BAIR introduces GRASP, a gradient-based planner that enables robust long-horizon planning in learned world models by lifting trajectories into virtual states and isolating action gradients to avoid adversarial state-input sensitivities.
Hyatt Deploys ChatGPT Enterprise to Enhance Global Operations
Hyatt has integrated ChatGPT Enterprise, providing its global corporate and hotel workforce with access to GPT 5.4 and Codex to automate manual tasks and improve guest experiences.
Qwen3.6-Max-Preview release notes / what's new
Qwen3.6-Max-Preview is a proprietary preview model from Qwen that improves upon Qwen3.6-Plus in agentic coding, world knowledge, and instruction following.
OpenAI Codex Update April 2026
OpenAI has updated Codex to support background computer use, long-term task automation, and memory, expanding its utility from a code generator to a full software development lifecycle partner.
GPT-Rosalind: OpenAI's Frontier Reasoning Model for Life Sciences
OpenAI has introduced GPT-Rosalind, a reasoning model optimized for biology, drug discovery, and translational medicine to accelerate early-stage research workflows.
Hugging Face transformers-to-mlx Skill and Test Harness
Hugging Face has released a Skill and a non-agentic test harness to streamline the porting of language models from the transformers library to mlx-lm while maintaining high code quality and reviewer signal.
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers
Hugging Face published a guide on training and finetuning multimodal embedding and reranker models with Sentence Transformers, demonstrating how task-specific finetuning improves performance on retrieval tasks such as Visual Document Retrieval.
OpenAI Trusted Access for Cyber Program
OpenAI has launched Trusted Access for Cyber, a program providing scaled access to advanced cyber capabilities and $10 million in API credits to a diverse ecosystem of defenders to enhance global digital resilience.
Ecom-RLVE: Adaptive Verifiable Environments for E-Commerce Conversational Agents
Hugging Face introduces EcomRLVE-GYM, a framework for training e-commerce agents using eight verifiable, multi-turn environments with adaptive difficulty scaling to bridge the gap between conversational fluency and actual task completion.
Gemini 3.1 Flash TTS release notes / what's new
Google DeepMind has released Gemini 3.1 Flash TTS, a text-to-speech model featuring natural language audio tags for precise control over vocal style, pacing, and delivery across 70+ languages.
VAKRA Benchmark Analysis: Agent Reasoning, Tool Use, and Failure Modes
Hugging Face announced the VAKRA benchmark, a tool‑grounded, executable suite that evaluates AI agents on compositional reasoning across 8,000+ APIs and document sources, revealing widespread failures in tool selection, multi‑hop reasoning, and policy adherence.
OpenAI Agents SDK Update
OpenAI has released updated capabilities for the Agents SDK, introducing a model-native harness and native sandbox execution to improve agent reliability and performance in production environments.