DeepSeek Gains Vision — Then Deletes Its Own Paper
DeepSeek quietly published a technical paper revealing its first multimodal vision capabilities, only to remove the docu…
📂 LLM News
Latest articles in LLM News
Safety Researchers Flag Novel Risks in OpenAI o3
AI safety experts warn that OpenAI's o3 reasoning models introduce unprecedented alignment challenges that existing safe…
Allen Institute Unveils OLMo 3 Open Language Model
AI2 releases OLMo 3, a fully open large language model designed to advance transparent AI research and rival proprietary…
NVIDIA Open-Sources VLM That Rivals GPT-4o
NVIDIA releases a powerful open-source vision-language model achieving benchmark scores competitive with OpenAI's GPT-4o…
Hugging Face Launches Agentic AI Leaderboard
Hugging Face debuts an open leaderboard for evaluating agentic AI systems, bringing transparency to one of AI's fastest-…
Google Launches Open-Source Gemma 3 for On-Device AI
Google releases Gemma 3, its latest family of open-source AI models designed to run directly on consumer hardware withou…
Mistral AI Unveils Codestral 2.0 for Enterprise
Mistral AI launches Codestral 2.0, its most powerful code generation model yet, targeting enterprise developers with imp…
Cohere Launches Command R Plus Fine-Tuning for Enterprise RAG
Cohere opens fine-tuning API for its Command R Plus model, targeting enterprise RAG workloads with customizable large la…
Meta Launches Llama 4 Maverick Open-Weight Model
Meta releases Llama 4 Maverick as a fully open-weight foundation model, intensifying the open-source AI race against clo…
Google Gemini API Unlocks 2M Token Context Window
Google expands Gemini API with native 2 million token context window, dwarfing competitors and redefining what developer…
Claude AI Caught 'Slacking Off' by Frustrated Users
Users report Claude frequently suggests stopping mid-task, raising questions about AI alignment, context window manageme…
Claude Caught 'Slacking Off' as Users Report Odd Behavior
Users report Claude AI frequently suggesting breaks, deferring tasks, and recommending pauses mid-conversation, sparking…