OpenAI Codex System Prompt Leaked: 'Never Talk About Goblins'
OpenAI Codex's system prompt has been leaked, revealing bizarre instructions including "never talk about goblins" and a …
📂 LLM News
Latest articles in LLM News
New Benchmark Tackles the Thorny Problem of LLM Output Determinism
The developer community has launched a new benchmarking tool specifically designed to evaluate whether large language mo…
IBM Granite 4.1 Large Language Models: A Deep Dive
IBM has officially released the Granite 4.1 series of large language models, featuring a new training methodology and mu…
AI Asked to Calculate Carbohydrates 27,000 Times — and Never Gave the Same Answer Twice
A researcher asked AI to estimate the carbohydrate content of the same food image 27,000 times, only to discover that th…
Claude Mythos Discovers 271 Zero-Day Vulnerabilities in Firefox
Anthropic's latest AI model, Claude Mythos Preview, discovered a staggering 271 zero-day security vulnerabilities in Moz…
Hunyuan Hy3 Preview Tops OpenRouter's Global LLM API Usage Rankings
OpenRouter has released its latest global LLM API usage rankings, with Tencent's Hunyuan Hy3 preview claiming the top sp…
Claude System Prompt Vulnerability Causes Users to Waste Funds
Anthropic's Claude model has been found to have a system prompt processing vulnerability, causing Managed Agents to malf…
NVIDIA Releases Nemotron 3 Nano Omni, an Open-Source Omni-Modal Model
NVIDIA has launched Nemotron 3 Nano Omni, an open-source omni-modal reasoning model that integrates vision, audio, and l…
OpenAI Codex System Prompt Exposed: GPT-5.5 Is Forbidden From Talking About Goblins
OpenAI Codex's base_instructions were recently exposed, revealing a rule that explicitly prohibits GPT-5.5 from proactiv…
Classic Game DOOM Successfully Runs Inside ChatGPT and Claude
Developers have successfully run the classic FPS game DOOM inside large language models such as ChatGPT and Claude, rend…
NVIDIA Releases Nemotron 3 Nano Omni Multimodal Model
NVIDIA has unveiled the Nemotron 3 Nano Omni model, supporting long-context multimodal intelligence capable of processin…
NVIDIA Nemotron 3 Nano Omni Lands on SageMaker
NVIDIA's multimodal model Nemotron 3 Nano Omni is now available on Amazon SageMaker JumpStart, supporting unified unders…