Home · 2
IBM and Hugging Face Release Granite 4.1 Preview Models Trained on LLMON Markup
Granite 4.1 3B and 8B LLMON preview models use special-token markup to separate prompt instructions from data and block indirect injections.
llama.cpp and ggml Reach v1.0.0, Formalizing API Stability and Semantic Versioning
llama.cpp and ggml release version 1.0.0, establishing stable C/C++ APIs, formal Semantic Versioning, and expanded cross-platform packaging.
Anthropic Updates Claude Code v2.1.323 with Experimental Code Mode and Plugin Versioning
Claude Code v2.1.323 introduces experimental Code Mode, plugin engine constraints, long-hook heartbeats, and memory bug fixes.
n8n Adds Native Public API Rate Limits and Safe Retry Mechanics in Version 2.47.0
n8n 2.47.0 introduces API key rate limits, standard 429 retry headers, Redis sync for multi-main setups, and 12-hour idempotency protection.
Claude Code v2.1.321 Adds Fable 5.2, Remote Control Settings, and Network Edit Fixes
Anthropic updates Claude Code to v2.1.321, adding Claude Fable 5.2, remote control settings flags, sandbox security patches, and network path fixes.
Google Deploys Gemini 4 Hydrogen Model Globally to Google Search
Google has deployed Gemini 4 Hydrogen globally in Search, updating query fan-out reasoning, instruction following, and structured visual responses.
Google Cloud Launches Gemini for Healthcare in Preview
Google Cloud releases Gemini for healthcare to automate prior authorizations and medical record synthesis across fragmented clinical systems.
Introducing the Gemini Decisions API and Gemini Decision 1
Google launches a new endpoint for fast, calibrated multimodal scoring. Evaluate up to 1,000 bounded choices in a single pass at $0.10/1M input tokens.
Tether AI Releases BotPsy: A 0.8B On-Device Agentic Model Built for Function Calling
Tether AI Research launched BotPsy, an open-source 0.8B parameter agent trained on top of Qwen3.5-0.8B for efficient on-device tool use.
n8n Adds Google Gemini 4 Carbon with High Security Benchmarks and Extended Execution
n8n integrates Google Gemini 4 Carbon, bringing strong prompt injection defense and multi-step agent execution to workflow automation.