OpenAI previews GPT-5.6 Sol, a next-generation model with enhanced capabilities in coding, science, and cybersecurity, along with an advanced safety stack.
OpenAI published a research paper about how AI agents are transforming work by enabling longer, more complex tasks and expanding productivity across roles.
OpenAI and Broadcom introduce a custom AI chip named Jalapeño, built for LLM inference to improve performance, efficiency, and scale across AI systems.
OpenAI is helping to build shared standards for advanced AI by supporting evaluation frameworks, safety practices, and global cooperation through the Appia Foundation.
OpenAI introduces new Daybreak tools, including Codex Security and GPT-5.5-Cyber, to help organizations find, validate, and patch vulnerabilities at scale.
OpenAI introduces Patch the Planet, a Daybreak initiative that uses AI and expert review to help open-source maintainers find, validate, and fix vulnerabilities.
Hugging Face announces that users can now spin up a private, OpenAI-compatible LLM endpoint on HF Jobs with a single command, using vLLM, with pay-per-second billing and no server provisioning.
NVIDIA announced NeMo AutoModel, an open library built on Transformers v5 that provides Expert Parallelism, DeepEP fused all-to-all dispatch, and TransformerEngine kernels to accelerate fine-tuning…
Hugging Face and Treble Technologies launch the FFASR Leaderboard, the first open, community-driven benchmark for evaluating ASR models under realistic far-field acoustic conditions, using hybrid…
This guest post by Thomas Steiner from Google Chrome experiments with the proposed Cross-Origin Storage API to address the challenge of redundant caching of AI model resources and WebAssembly runtime…
Hugging Face describes a new weekly release pipeline for huggingface_hub that uses an open-weights model (GLM-5.2) to draft release notes and Slack announcements, with a human-in-the-loop for review.
PP-OCRv6 is the latest generation of PaddleOCR's OCR model family, with three tiers (tiny, small, medium) supporting up to 50 languages, using PPLCNetV4 backbone, RepLKFPN for detection, and…
Hugging Face blog post describes using local models Gemma and Qwen in an agent harness to triage issues and pull requests in the OpenClaw repository, demonstrating a real-time notification system…
Google Research introduces a method to retrofit Multi-Token Prediction onto frozen Gemini Nano v3 models, accelerating on-device inference on Pixel 9 and 10 series without separate drafters.
Google Research announces linear elastic caching, a new approach to cache management that treats memory as a variable cost and uses a ski rental formulation to dynamically adjust cache size.
Google Research presents a study on how reasoning helps LLMs recall simple facts even when step-by-step reasoning is unnecessary, identifying two mechanisms: a computational buffer effect and factual…
Meta presents a case study on its hybrid approach to privacy-aware infrastructure (PAI) for asset classification, combining LLM-based reasoning with deterministic, human-reviewed rules to handle…
Meta engineers developed ultra-narrow steel-can batteries for smart glasses, achieving widths as narrow as 7mm, and improved capacity across generations, including a 248 mAh cell for display glasses.
Meta describes its multi-year effort to adopt the AV1 video codec for real-time communication, covering codec selection, device eligibility, rate control, error resilience, and the challenges of…
Microsoft Research, in collaboration with UC Berkeley, UCSF, and Columbia University, published a paper in Nature Neuroscience introducing generative causal testing (GCT), a framework that distills…
Talos is an open-source tool for automated, iterative reanalysis of genomic data to diagnose rare diseases, designed to be conservative with low false-positive rates, recovering 90% of diagnoses in…
Mistral AI announced a new feature providing more control over connectors, enabling users to better manage integrations or connections within their platform.
Interactive Brokers integrates with Grok, adding AI-powered portfolio analysis, scenario modeling, research, and order instructions to the trading platform.
Alibaba releases Qwen-AgentWorld, a native language world model for simulating agent environments across seven domains, along with the AgentWorldBench benchmark.
ByteDance announces the official release of the Seed2.1 model family, a new generation of agent-capable models with enhanced general agent capabilities, end-to-end coding delivery, and stronger…
Cursor released a new Customize page that unifies plugins, skills, MCPs, subagents, rules, commands, and hooks into one place, and added a marketplace leaderboard, plugin canvases, and team…
LG AI Research proposes a cluster-aware upcycling method that initializes experts and routers based on activation clustering of a pre-trained dense model, aiming to improve expert specialization in…
Together AI introduces ParallelKernelBench (PKB), a benchmark and evaluation framework for multi-GPU kernel generation, containing 87 problems from real codebases.