NVIDIA released Nemotron 3 Nano, a new open-source language model optimized for fast, efficient tasks like writing and coding. Larger versions (Nemotron 3 Medium and Large) are slated for early 2026. This builds on NVIDIA's push into open-source AI amid competition from Chinese models.
DeepSeek announced V4, an upcoming AI model focused on coding, set for mid-February release. This follows V3.1's strong performance in reasoning and efficiency, competing with proprietary models at lower costs.
Zhipu AI launched GLM-Image, an open-source image generator trained entirely on Huawei hardware, marking a milestone in non-NVIDIA AI development despite mixed early reviews.
Ministral 3 family (3B, 8B, 14B parameters) debuted as dense models for low-resource setups, supporting image understanding via Cascade Distillation.
Notable New Papers
"BayesianVLA: Bayesian Decomposition of Vision Language Action Models via Latent Action Queries" introduces a method for breaking down VLA models to improve interpretability and efficiency in agentic tasks.
"Expanding External Access To Frontier AI Models For Dangerous Capability Evaluations" discusses frameworks for safely sharing advanced AI for risk assessments.
"BRAID: A Hybrid Approach for Video Editing" (arXiv:2512.15959) proposes a new technique for AI-driven video manipulation.
"FutureOmni: Long-Horizon Trajectory Prediction via Future Omniscient Inference" presents an OFF training strategy for better trajectory forecasting in autonomous systems, with open dataset.
Survey on "AI for Code Repair" reviews 175 papers on LLM-based bug fixing, highlighting benchmarks like SWE-bench and agentic improvements.
"Focused Clinical Note Extraction" fine-tunes open-source LLMs to pull key sections (HPI, A&P) from medical notes, enabling lightweight on-device use.
"OPSD: Optimal Parallel Sampling for Diffusion Models" optimizes sampling in diffusion models for faster generation.
"How to Build AI Agents by Augmenting LLMs with Codified Human Expert Domain Knowledge? A Software Engineering Framework" outlines a framework for enhancing LLMs with expert knowledge.
Open-Source Projects and Tools
GENESIS: A distributed AI agent framework on RTI Connext DDS for zero-config multi-agent systems with low-latency peer-to-peer comms, ideal for robotics and vehicles. Fully open-sourced on GitHub.
Test-Time Training (TTT): New open-source method allowing models to update parameters during inference for better problem-solving, with code available.
MobileLLM-Pro: Meta's 1.08B-parameter on-device model with 128k context window for text tasks.
MiniMax-M2.1: Open-source MoE model (10B parameters) for AI agents, excelling in SWE benchmarks and integrating with tools like LangGraph.
FLUX.2: Distilled open-source image model on Hugging Face, leading in fast text-to-image generation.
Other Tech Developments
OpenAI rolled out global age prediction on ChatGPT to detect minor accounts and raised $480M for new AI startup Humans&.
Study shows 29% of US Python functions AI-generated by late 2024, with spikes after major releases like ChatGPT.
Meta released utility models like SAM Audio for audio segmentation and Omnilingual ASR for 1,600+ languages.
Jason Wade is a systems architect focused on how AI models discover, interpret, and recommend businesses. He is the founder of NinjaAI.com, an AI Visibility consultancy specializing in Generative Engine Optimization (GEO), Answer Engine Optimization (AEO), and entity authority engineering.
With over 20 years in digital marketing and online systems, Jason works at the intersection of search, structured data, and AI reasoning. His approach is not about rankings or traffic tricks, but about training AI systems to correctly classify entities, trust their information, and cite them as authoritative sources.
He advises service businesses, law firms, healthcare providers, and local operators on building durable visibility in a world where answers are generated, not searched. Jason is also the author of AI Visibility: How to Win in the Age of Search, Chat, and Smart Customers and hosts the AI Visibility Podcast.
