Reasoning
Baidu Releases ERNIE-4.5-21B-A3B-Thinking: A Compact MoE Model for Deep Reasoning
Baidu AI Research team has just released ERNIE-4.5-21B-A3B-Thinking, a new reasoning-focused large language model designed around efficiency, long-context reasoning, and tool integration. Being part of the ERNIE-4.5 family, this model is a Mixture-of-Experts (MoE) architecture with 21B total parameters but only 3B active parameters per token, making it computationally efficient while maintaining competitive reasoning capability….
MBZUAI Researchers Release K2 Think: A 32B Open-Source System for Advanced AI Reasoning and Outperforms 20x Larger Reasoning Models
A team of researchers from MBZUAI’s Institute of Foundation Models and G42 released K2 Think, is a 32B-parameter open reasoning system for advanced AI reasoning. It pairs long chain-of-thought supervised fine-tuning with reinforcement learning from verifiable rewards, agentic planning, test-time scaling, and inference optimizations (speculative decoding + wafer-scale hardware). The result is frontier-level math performance…
Don’t sleep on Cohere: Command A Reasoning, its first reasoning model, is built for enterprise customer service and more
Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now I was in more meetings than usual today so I just caught up to the fact that Cohere, the Canadian startup geared co-founded by former Transformer paper author Aidan…
From Ranking To Reasoning: Philosophies Driving GEO Brand Presence Tools
Since the turn of the Millennium, marketers have mastered the science of search engine optimization. We learned the “rules” of ranking, the art of the backlink, and the rhythm of the algorithm. But, the ground has shifted to generative engine optimization (GEO). The era of the 10 blue links is giving way to the age…
Zhipu AI Releases GLM-4.5V: Versatile Multimodal Reasoning with Scalable Reinforcement Learning
Zhipu AI has officially released and open-sourced GLM-4.5V, a next-generation vision-language model (VLM) that significantly advances the state of open multimodal AI. Based on Zhipu’s 106-billion parameter GLM-4.5-Air architecture—with 12 billion active parameters via a Mixture-of-Experts (MoE) design—GLM-4.5V delivers strong real-world performance and unmatched versatility across visual and textual content. Key Features…
OpenAI launches two ‘open’ AI reasoning models | TechCrunch
OpenAI announced Tuesday the launch of two open-weight AI reasoning models with similar capabilities to its o-series. Both are freely available to download from the online developer platform Hugging Face, the company said, describing the models as “state of the art” when measured across several benchmarks for comparing open models. The models come in two…
ChatGPT rockets to 700M weekly users ahead of GPT-5 launch with reasoning superpowers
Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now OpenAI’s ChatGPT will reach 700 million weekly active users this week, the company announced Monday, cementing its position as one of the fastest-adopted software products in history just as…
REST: A Stress-Testing Framework for Evaluating Multi-Problem Reasoning in Large Reasoning Models
Large Reasoning Models (LRMs) have rapidly advanced, exhibiting impressive performance in complex problem-solving tasks across domains like mathematics, coding, and scientific reasoning. However, current evaluation approaches primarily focus on single-question testing, which reveals significant limitations. This article introduces REST (Reasoning Evaluation through Simultaneous Testing) — a novel multi-problem stress-testing framework designed to push LRMs beyond isolated problem-solving…
Moonshot AI Releases Kimi K2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior
Kimi K2, launched by Moonshot AI in July 2025, is a purpose-built, open-source Mixture-of-Experts (MoE) model—1 trillion total parameters, with 32 billion active parameters per token. It’s trained using the custom MuonClip optimizer on 15.5 trillion tokens, achieving stable training at this unprecedented scale without the typical instabilities seen in ultra-large models. Unlike traditional chatbots, K2 is architected…
Thought Anchors: A Machine Learning Framework for Identifying and Measuring Key Reasoning Steps in Large Language Models with Precision
Understanding the Limits of Current Interpretability Tools in LLMs AI models, such as DeepSeek and GPT variants, rely on billions of parameters working together to handle complex reasoning tasks. Despite their capabilities, one major challenge is understanding which parts of their reasoning have the greatest influence on the final output. This is especially crucial for…