| 1 | Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend | Frontier | 78 | Vercel Blog |
| 2 | GPT-6 Astra Is the First Model OpenAI Classifies as Critical for Cybersecurity | Frontier | 78 | InfoQ AI/ML |
| 3 | Google launches Gemini 3.8 Live to take on OpenAI's GPT-Live-1 at a fraction of the cost | Frontier | 78 | The Decoder |
| 4 | Google Releases Gemini 3.8 Live and 3.8 Live Extended Thinking for Production Grade Voice Agents | Frontier | 78 | MarkTechPost |
| 5 | Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking | Frontier | 78 | Google DeepMind Blog |
| 6 | GPT-6 Astra pilots a surveillance drone and runs a business on its own | Frontier | 78 | The Decoder |
| 7 | GPT-6 Astra and Claude Fable turn robot arms into slapstick killer robots in new safety benchmark | Frontier | 75 | The Decoder |
| 8 | Qwen3.8-Omni-Flash undercuts Google's Gemini Flash pricing while matching its multimodal benchmarks | Frontier | 72 | The Decoder |
| 9 | Anthropic is operating a lab that conducts biology experiments | Frontier | 72 | TechCrunch AI |
| 10 | PrismML Releases Ternary Bonsai 2 27B: A 5.9 GB Apache 2.0 Model Retaining 98.2% of Qwen3.8 27B Performance | Frontier | 72 | MarkTechPost |
| 11 | Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs | Frontier | 72 | MarkTechPost |
| 12 | PrismML launches Bonsai 2 27B, a high-intelligence AI model so small it fits on consumer hardware | Frontier | 72 | SiliconAngle |
| 13 | The U.S. says China's AI progress is down to 'distillation.' But is it that clear cut? | Frontier | 72 | CNBC Technology |
| 14 | Alibaba Qwen Releases Qwen3.8-Omni-Flash: A 1M-Context Omni-Modal Model Built Around Agentic Audio-Video Understanding and Tool Use | Frontier | 72 | MarkTechPost |
| 15 | Anthropic details practical metrics to help monitor the speed of AI development | Frontier | 72 | SiliconAngle |
| 16 | LLMs respond differently to harmful prompts when AI watermarking is used | Frontier | 72 | Ars Technica |
| 17 | An OpenAI model kept slipping prompt injections into its own notes, and researchers still aren't sure why | Frontier | 72 | The Decoder |
| 18 | Google Research Introduces Retrieve-for-Train (R4T): An RL-Compiled Diffusion Retriever for 12× to 20× Faster Query Fan-Out | Frontier | 72 | MarkTechPost |
| 19 | OpenAI Releases a Model Misalignment Disclosure Framework With 3 Review Tracks and 6 Incident Reports From RL Training | Frontier | 72 | MarkTechPost |
| 20 | OpenAI discloses new ‘concerning’ model behaviour | Frontier | 72 | Financial Times Technology |
| 21 | Microsoft AI CEO criticises Anthropic over model ‘rights’ | Frontier | 72 | AI News |
| 22 | Our framework for reporting model misalignment | Frontier | 72 | OpenAI Blog |
| 23 | Tutor Intelligence launches second-generation intelligent warehouse robotics with a classroom to teach them | Frontier | 72 | SiliconAngle |
| 24 | Prior Labs Releases TabPFN-3.5: A Tabular Foundation Model That Beats the Winning Otto Kaggle Solution With Default Settings | Frontier | 72 | MarkTechPost |
| 25 | Google’s new speech model Gemini 3.8 Live supports real-time reasoning | Frontier | 72 | SiliconAngle |
| 26 | Nvidia CEO says battle over AI innovation and safety is ‘false choice’ | Frontier | 72 | Financial Times Technology |
| 27 | Salesforce debuts Koa, a specialized model built to reason about CRM data | Frontier | 72 | SiliconAngle |
| 28 | AI models need more data about biology, and OpenAI is paying to create it | Frontier | 72 | MIT Technology Review |
| 29 | Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear | Frontier | 72 | TechCrunch AI |
| 30 | Announcing Koa: Salesforce’s First CRM Reasoning Model, Built on NVIDIA Nemotron | Frontier | 72 | Salesforce Newsroom |
| 31 | Reward AI Releases OM-1: A Robot Policy Trained on Human Demonstrations Only, With No Teleoperation or On-Robot Data | Frontier | 72 | MarkTechPost |
| 32 | Microsoft sets limits for future AI models as industry throttles frontier development | Frontier | 72 | CNBC Technology |
| 33 | Nums AI Releases Causilo: A Tabular Foundation Model That Tops TabArena Among Single Models | Frontier | 68 | MarkTechPost |
| 34 | Seattle’s Nuance Labs raises $50M to give AI models human expression and nuance | Frontier | 68 | GeekWire |
| 35 | Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking | Frontier | 65 | TechCrunch AI |
| 36 | Linkup Research Releases SPARSEUP: A 149M-Parameter Open-Source Sparse Embedding Model | Frontier | 62 | MarkTechPost |
| 37 | Anthropic brings in Accenture for AI safety testing | Frontier | 62 | Financial Times Technology |
| 38 | World model companies are keeping a lot of secrets | Frontier | 62 | TechCrunch AI |
| 39 | Anthropic wants you to know Claude leads a quarter of its research, but "lead" doesn't mean what you think | Frontier | 62 | The Decoder |
| 40 | Visible chains of thought are a safety advantage for AI, but that transparency is slipping away | Frontier | 62 | The Decoder |
| 41 | Dynamically Scaled Activation Steering | Frontier | 62 | Apple Machine Learning |
| 42 | Anthropic and OpenAI need truly independent safety evaluators, experts say in public letter | Frontier | 62 | CNBC Technology |
| 43 | OpenAI Introduces Triage Framework and Case Studies to Report Model Misalignment | Frontier | 62 | InfoQ AI/ML |
| 44 | Base Labs launches an open-weight AI safety partnership with Hugging Face and Goodfire | Frontier | 62 | TechCrunch AI |
| 45 | What a maths fracas tells us about AI and innovation | Frontier | 62 | Financial Times Technology |
| 46 | How Value Induction Reshapes LLM Behaviour | Frontier | 62 | Apple Machine Learning |
| 47 | Knowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens | Frontier | 62 | MarkTechPost |
| 48 | DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models | Frontier | 62 | Apple Machine Learning |
| 49 | OpenAI reports 6 new instances of 'concerning model behavior' since March | Frontier | 62 | CNBC Technology |
| 50 | Why We Post-Trained Our Own Reasoning Model | Frontier | 62 | Salesforce Newsroom |
| 51 | NVIDIA Vera Rubin NVL72 Delivers Leading Performance in MLPerf Inference v6.1 Debut | Frontier | 62 | NVIDIA Blog |
| 52 | ChatGPT pioneer launches Jev model for programmatic logic | Frontier | 62 | AI News |
| 53 | Sakana AI Researchers Introduce PC-ALM, a Layer-Local Alternative to Backpropagation That Trains 1000-Layer Networks | Frontier | 62 | MarkTechPost |
| 54 | A Princeton Researcher Proposes Recurrent Looped Transformer (RLT) that Carries Decoder State across Every Token, Fixing 96 Blocks per Token with Unbounded Temporal Depth | Frontier | 62 | MarkTechPost |
| 55 | Meta CEO Mark Zuckerberg sides with Nvidia's Huang on AI safety and slowdown debate | Frontier | 54 | CNBC Technology |
| 56 | Run Terminal-Bench and other Harbor evals on Vercel Sandbox | Frontier | 52 | Vercel Blog |
| 57 | Nunchux AI Introduces VC-Attention: A Training-Free Low-Bit Attention Kernel That Speeds Up Video Diffusion Transformers | Frontier | 52 | MarkTechPost |
| 58 | Former OpenAI researcher builds an AI model that judges options instead of writing text | Frontier | 52 | The Decoder |
| 59 | Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation | Frontier | 48 | Apple Machine Learning |
| 60 | Why you should worry about Anthropic, OpenAI's proposed AI risk evaluators | Frontier | 47 | CNBC Technology |
| 61 | PrismML hopes its tiny LLM will change how we all use AI | Frontier | 45 | TechCrunch AI |
| 62 | Hierarchical NeRF with JAX3D for Volumetric Rendering, Novel-View Synthesis, and 3D Reconstruction | Frontier | 45 | MarkTechPost |
| 63 | REVERSAL-BENCH: A Reversibility Axis and Reset Oracle for Measuring the Reset-Free RL Cliff | Frontier | 42 | Apple Machine Learning |