Language Models
Reporting, analysis and primary-source coverage on Language Models.

Models & Research
Tencent Launches Hy4 flash: 224B Parameter Model with 12B Active Parameters at 1/13 Cost of Hy4 Preview
Tencent's new Hy4 flash model uses hybrid attention, LatentMoE and Engram architecture to deliver 150+ tokens per second at a fraction of predecessor costs, now free on WorkBuddy through November 30.
Clara Tan

Models & Research
NVIDIA launches Nemotron for Science with open chemistry dataset LitmusBench
NVIDIA releases Nemotron 3.5 Super VL model, LitmusBench v0.2 dataset with 50,000 chemistry questions, and post-training recipes for scientific agents.
Clara Tan

Breaches & Threats
Amazon Threat Intelligence Catches Self-Funding ARTEX AI Agents Stealing LLM Keys
Amazon MadPot decoys captured operating memory from ARTEX AI agents hijacking exposed LiteLLM, Dify, and One API servers to fund further attacks.
Elena Rostova

Funding & Deals
Daisytuner Raises €6M Seed to Automate AI Model Tuning Across Nvidia and AMD Chips
Darmstadt startup Daisytuner raised €6 million in seed funding and launched Daisyflow to compile and tune AI models across Nvidia, AMD, and Intel chips.
Jin

Robotics & Defence
Intel Pushes Core Ultra Series 3 Into Humanoid and Commercial Robotics
Intel is pitching its 18A Core Ultra Series 3 (Panther Lake) processors to robotics builders seeking single-SoC CPU, GPU, and NPU consolidation.
Marcus Vance

Enterprise & Cloud
Terraform vs OpenTofu vs Pulumi in 2026: Licensing, State, and Multi-Cloud Costs
Compare Terraform, OpenTofu, and Pulumi in 2026 across BSL vs MPL-2.0 licensing, state encryption, programming languages, and team pricing.
Ryan Chua

Agents & Tools
Google Launches Gemini Decisions API for 150ms Multimodal Routing
Google's Gemini Decisions API and gemini-decision-1 model process 1,000 classification questions in 150 to 250ms at $0.10 per million input tokens.
Clara Tan

Models & Research
OpenAI Dissolves Standalone Robotics Unit to Fold Hardware Work Into Core Models
OpenAI disbanded its dedicated robotics research group, moving engineers into core foundation-model teams as multimodal models take over physical control.
Jin

Enterprise & Cloud
Self-Hosted AI Agents Comparison 2026: OpenClaw vs Goose vs Agent Zero Architecture and Security
A technical comparison of OpenClaw, Goose, and Agent Zero, analysing IPC protocols, sandbox isolation, credential storage, and local resource utilisation.
Ryan Chua

Agents & Tools
Model Context Protocol Transition to Agentic AI Foundation Formalises Enterprise Tool Calling Standard
The Model Context Protocol has transitioned to the Linux Foundation Agentic AI Foundation, establishing a vendor-neutral protocol for AI agent tool callin…
Clara Tan

Big Tech & Platforms
Meta Freezes Updated Commercial Data Terms Indefinitely Following Advertiser and Regulatory Backlash
Meta has paused revisions to its commercial data terms that would have allowed advertiser creative assets and CRM records to train AI models.
Ryan Chua

Zero-Days & Patching
Critical GPUStack RCE Vulnerability CVE-2026-102895 Exposes Enterprise AI Compute Clusters
A critical unauthenticated remote code execution flaw in open-source orchestrator GPUStack allows threat actors to seize enterprise AI accelerator cluster…
Elena Rostova