The archive · 11 labs · 3,050 dispatches

The labs

No more opening a dozen official blogs every morning. First-hand releases from OpenAI, Anthropic, DeepMind and the rest, each with its substance pulled out.

401

NVIDIA NeMo AutoModel Accelerates Fine-Tuning of Mixture-of-Experts Models

NVIDIA NeMo AutoModel provides 3.4-3.7x higher training throughput and 29-32% lower GPU memory usage for fine-tuning Mixture-of-Experts models by building on Hugging Face Transformers v5 with Expert Parallelism, DeepEP, and TransformerEngine kernels, requiring only a one-line import change.

402

Mistral AI introduces granular admin controls, scoped API keys, and multi‑account connectors

Mistral AI announced new connector features—including GA admin controls, scoped API keys, multi‑account connectors, a Debugger, and integrations in Workflows and Vibe Code—to improve security and reliability for enterprise AI workloads.

403

OpenAI and Broadcom Unveil Jalapeño LLM Inference Chip

OpenAI and Broadcom have introduced Jalapeño, a custom-designed AI accelerator optimized specifically for LLM inference to improve performance per watt and reduce compute costs.

404

Hugging Face FFASR Leaderboard: Benchmarking Far-Field ASR

Hugging Face and Treble Technologies have launched the FFASR Leaderboard, the first open community-driven benchmark to quantify the performance gap between near-field and far-field Automatic Speech Recognition (ASR) in realistic acoustic environments.

405

GPT-5 Pro in Immunology: Solving T-Cell Specialization Mysteries

Immunologist Derya Unutmaz used GPT-5 Pro to solve a three-year-old mystery regarding how deoxyglucose affects T-cell specialization, demonstrating the model's ability to generate novel biological insights and predict unpublished experimental outcomes.

406

OpenAI and the Appia Foundation: Establishing Shared Standards for Advanced AI

OpenAI has helped found the Appia Foundation to develop open, modular specifications that translate international AI standards into practical assessment criteria to ensure interoperability across organizations and jurisdictions.

407

Mistral OCR 4 release notes / what's new

Mistral AI has released Mistral OCR 4, a high-performance document parsing model supporting 170 languages and providing structured outputs including bounding boxes, block classification, and confidence scores.

408

Qwen-AgentWorld release: language world model for seven domains and its impact on general agents

Qwen releases Qwen‑AgentWorld, a language world model that simulates seven agent environments and improves general agents via controllable simulation and unified next‑state prediction.

409

vLLM-Omni TTS Inference Engineering

vLLM-Omni engineered TTS inference for models like Qwen3-TTS, VoxCPM2, Higgs Audio V3, and Fish Speech S2 Pro by applying model-specific optimizations that decouple latency and throughput bottlenecks, significantly improving audio throughput and reducing end-to-end latency.

410

Experimenting with the Cross-Origin Storage API in Transformers.js

Hugging Face shows how Transformers.js can use the experimental Cross-Origin Storage API to cache model and Wasm resources by hash, eliminating duplicate downloads across origins.

411

Omio Conversational Travel and AI-Native Operations

Omio is leveraging OpenAI models to transition from search-based travel planning to conversational commerce and reducing product development time to approximately 20% of previous levels.

412

Hugging Face huggingface_hub Release Automation

Hugging Face has transitioned from a 4-6 week release cycle to a weekly cadence for huggingface_hub by implementing an AI-driven, human-in-the-loop CI/CD pipeline using open-source tools and open-weights models.

413

PP-OCRv6 release: 50-Language OCR from 1.5M to 34.5M Parameters

PaddlePaddle has released PP-OCRv6, a scalable OCR model family supporting 50 languages with parameter counts ranging from 1.5M to 34.5M.

414

Daybreak: Tools for securing every organization in the world

OpenAI announced an expansion of Daybreak, releasing an updated Codex Security plugin, the full version of GPT‑5.5‑Cyber, a cyber partner program, and the Patch the Planet initiative to help defenders find, validate, and patch vulnerabilities at machine speed.

415

OpenAI Patch the Planet Initiative

OpenAI has launched Patch the Planet, a Daybreak initiative in partnership with Trail of Bits to use AI-assisted security research and human review to identify and patch vulnerabilities in critical open-source software.

416

Hugging Face Local Models for OpenClaw PR Triage

Hugging Face demonstrates how local models like Gemma 4 and Qwen 3.6, deployed in an agentic harness, can perform real-time, cost-free triage of GitHub issues and pull requests for the OpenClaw repository.

417

OpenAI Codex-maxxing for long-running work

OpenAI has released a whitepaper detailing strategies for using Codex as a persistent workspace to manage complex, multi-prompt workflows and long-running projects.

418

Samsung Electronics Deployment of ChatGPT Enterprise and Codex

Samsung Electronics is deploying ChatGPT Enterprise and Codex to all employees in Korea and all Device eXperience (DX) employees worldwide to enhance productivity across R&D, manufacturing, and corporate functions.

419

MosaicLeaks: Addressing Privacy Leakage in Deep Research Agents

Hugging Face introduces MosaicLeaks, a benchmark and the Privacy-Aware Deep Research (PA-DR) training method to prevent research agents from leaking private enterprise data through their external web queries.

420

ChatGPT Enterprise Usage Analytics and Spend Controls Update

OpenAI has introduced credit usage analytics and updated spend controls for ChatGPT Enterprise to help organizations track credit consumption and manage AI costs at scale.

421

Improving Health Intelligence in ChatGPT

OpenAI has updated ChatGPT with GPT-5.5 Instant, which demonstrates health intelligence comparable to frontier Thinking models and a 71% reduction in factuality issues in production health traffic.

422

OpenAI o3 Deep Research for Rare Genetic Disease Diagnosis

Researchers used OpenAI o3 Deep Research to achieve a 4.8% additional diagnostic yield in 376 unsolved rare childhood genetic disease cases through an AI-assisted research workflow.

423

Hugging Face Benchmarking Open Models on Agentic Tooling

Hugging Face introduces a new benchmarking harness to evaluate how different model sizes and library revisions affect the efficiency and success rate of coding agents using software tools.

424

Hugging Face PEFT: Evaluating Alternatives to LoRA

Hugging Face's benchmarking of the PEFT library reveals that while LoRA is widely popular, other parameter-efficient fine-tuning techniques like OFT and Lily can outperform it in memory efficiency and test accuracy depending on the task.

425

Anthropic Project Fetch Phase Two

Claude Opus 4.7 demonstrated the ability to autonomously perform robotic control tasks approximately 20 times faster than human teams, signaling a shift toward physical agentic AI.

426

Grok for Word Release Notes

xAI has released Grok for Word, a free Microsoft 365 add-in that enables users to generate structured documents, perform web and X research, and integrate data from external connectors.

427

Grok Models Now Available on Databricks Agent Bricks

xAI has integrated Grok models into the Databricks Agent Bricks platform, allowing enterprises to build AI agents that operate on Lakehouse data with governed access.

428

Strands Robots and LeRobot Integration: From Hugging Face Hub Datasets to Physical Robot Deployment

Hugging Face announced the Strands Robots SDK integration with LeRobot, enabling users to record robot demonstrations, push them to the Hub, run policies in simulation, and deploy the same code to physical SO-101 robots with a single argument change, while coordinating multiple robots via a Zenoh-based mesh.

429

OpenAI AI Chemist: Improving Chan-Lam Coupling with GPT-5.4

OpenAI and Molecule.one used GPT-5.4 and the Maria AI agent to autonomously identify and validate a method for improving the yield of primary sulfonamide Chan-Lam coupling reactions using TEMPO.

430

GLM-5.2: Built for Long-Horizon Tasks

GLM-5.2, released by Z.AI on Hugging Face, introduces a solid 1M-token context, improved architecture via IndexShare and MTP enhancements, effort-level control, and strong open-source performance on long-horizon coding benchmarks.

431

Agentic Resource Discovery (ARD) Specification and Hugging Face Implementation

Hugging Face has launched a reference implementation of the Agentic Resource Discovery (ARD) specification, an open standard that allows AI agents to dynamically search for and integrate tools, skills, and other agents at runtime.

432

OpenAI Introduces LifeSciBench for Evaluating Agentic AI in Life Sciences

OpenAI has released LifeSciBench, a benchmark of 750 expert-authored tasks designed to measure how AI systems handle complex, real-world life science research workflows rather than simple fact recall.

433

Anthropic Opens Seoul Office to Expand Korean AI Ecosystem

Anthropic has opened a new office in Seoul and signed a Memorandum of Understanding with Korea's Ministry of Science and ICT to advance AI safety and expand Claude's adoption across Korean enterprises and research institutions.

434

Grok 4.3 Release on Amazon Bedrock

xAI has made Grok 4.3 generally available on Amazon Bedrock, featuring a 1-million-token context window and the lowest hallucination rate among frontier models.

435

Google DeepMind AI-Accelerated Planning for UK House-Building

Google DeepMind is partnering with the UK government to develop a Gemini-powered AI planning prototype designed to halve the time it takes to process householder planning applications.

436

DeepMind AI Control Roadmap announcement

DeepMind released its AI Control Roadmap, a defense‑in‑depth framework that treats internal AI agents as insider threats and adds monitoring, supervision, and response layers to secure increasingly capable agents.

437

Qwen Robot Suite: Unified Foundation Models for Navigation, Manipulation, and World Modeling

Qwen introduced the Qwen‑Robot Suite—three foundation models (RobotNav, RobotManip, RobotWorld) that translate language into navigation, manipulation, and world‑prediction actions, enabling unified agentic robotics across dozens of embodiments.

438

vLLM Semantic Router Fusion primitive enables programmable multi‑model serving

vLLM introduced the Fusion primitive for its Semantic Router, enabling programmable multi‑model panels, judging, and synthesis as a first‑class serving pattern.

439

OpenAI Deployment Simulation: Pre‑release Risk Forecasting Using Real‑World Conversation Replay

OpenAI introduced Deployment Simulation, a method that replays real user conversations with a candidate model before release to predict undesirable behavior rates and improve safety assessments.

440

Qwen-RobotWorld: Boundless Worlds for Embodied Agents

Qwen-RobotWorld is a unified world model that uses natural language as a universal action interface to enable cross-scenario physical generalization across 20+ robot embodiments.

441

Qwen-RobotManip: Alignment Unlocks Scale for Robotic Manipulation Foundation Models

Qwen-RobotManip is a Vision-Language-Action (VLA) foundation model that uses a unified alignment framework and a human-to-robot data synthesis pipeline to achieve state-of-the-art generalization across diverse robot embodiments and out-of-distribution tasks.

442

Qwen-RobotNav: A Scalable Navigation Model for Agentic Systems

Qwen-RobotNav is a unified navigation model based on Qwen3-VL that achieves state-of-the-art performance across five navigation domains by treating visual context as a controllable inference-time interface.

443

Anthropic Claude Code Usage Report June 2026 – Findings on Agentic Coding, Expertise, and Labor Implications

Anthropic analyzed ~400,000 Claude Code sessions (Oct 2025‑Apr 2026) and found that users plan tasks while Claude executes them, success rates are comparable to software engineers across occupations, and domain expertise—not coding skill—drives higher success and value.

444

Grok for PowerPoint Release Notes

xAI has released Grok for PowerPoint, a free Microsoft 365 add-in that allows users to generate slide decks, research content, and apply styling using Grok's AI capabilities.

445

Grok Imagine Video 1.5 Release Notes

xAI has released Grok Imagine Video 1.5, an image-to-video model featuring improved motion, physics, and integrated audio generation with a faster generation speed.

446

Grok Build Agent Dashboard Release

xAI has introduced the Agent Dashboard in Grok Build, a centralized management interface that allows developers to monitor, dispatch, and interact with multiple parallel coding sessions simultaneously.

447

OpenAI Partner Network Announcement

OpenAI has launched the OpenAI Partner Network, a global ecosystem supported by a $150 million investment to help enterprises identify use cases, redesign workflows, and deploy AI solutions at scale.

448

OpenAI Academy Courses for AI Adoption at Work

OpenAI has launched three new courses—AI Foundations, Applied AI Foundations, and Agents and Workflows—to help organizations build AI fluency and transition from individual task improvement to structured, agent-assisted workflows.

449

MiniMax M3 vLLM Support: Day-0 Serving for 1M-Token Multimodal Reasoning

vLLM has released day-0 support for the MiniMax M3 model family, enabling efficient serving of 1M-token context and native multimodal reasoning using MiniMax Sparse Attention (MSA).

450

Preply and OpenAI: Personalizing Language Learning with AI

Preply has integrated OpenAI APIs to create Lesson Insights, a tool that transforms 1:1 human tutoring sessions into personalized learning journeys through automated feedback and targeted practice.