Google has launched Gemini 2.5 Pro, describing it as their “most intelligent AI model” to date. The new model represents a significant advancement in Google’s AI capabilities, with a particular focus on reasoning abilities that are now built directly into the system. According to Google’s announcement, Gemini 2.5 models are “thinking models” that can reason …
Tencent Holdings has introduced its Hunyuan T1 AI reasoning model that matches DeepSeek’s R1 in both performance and pricing. According to Coco Feng of the South China Morning Post, the model scored 87.2 points on the MMLU Pro benchmark, surpassing DeepSeek-R1’s 84 points but falling short of OpenAI’s o1 at 89.3 points. Pricing is competitive …
OpenAI has released o1-pro, an upgraded version of its o1 reasoning AI model, available to select developers who have spent at least $5 on OpenAI API services. As reported by Kyle Wiggers, o1-pro costs $150 per million tokens for input and $600 per million tokens for output, making it twice as expensive as GPT-4.5 for …
At the GPU Technology Conference (GTC) 2025, Nvidia announced a new family of AI models called Llama Nemotron designed to enhance reasoning capabilities for autonomous AI agents. These models are based on Meta’s open-source Llama models but have been refined through post-training optimization techniques to improve their performance in complex tasks such as multistep math, …
Baidu has released two new AI models, ERNIE 4.5 and ERNIE X1, claiming they outperform competitors like DeepSeek and OpenAI on various benchmarks while offering significantly lower pricing. Carl Franzen, writing for VentureBeat, reports that ERNIE 4.5 is a multimodal language model while X1 focuses on reasoning capabilities. The models are notably cheaper than competitors, …
Alibaba’s Qwen Team has introduced QwQ-32B, a new open-source language model that matches the performance of much larger models like DeepSeek-R1 despite having significantly fewer parameters. The 32-billion-parameter model, released under the Apache 2.0 license, leverages reinforcement learning (RL) to enhance reasoning capabilities for complex problem-solving tasks. Key features and capabilities QwQ-32B demonstrates impressive performance …
In a comprehensive analysis for Bloomberg, Azeem Azhar explores how artificial intelligence is fundamentally changing our economic model by making expertise abundant and affordable. Azhar, who writes the Exponential View newsletter, argues that AI is the next major step in the historical progression of lowering intelligence costs, following innovations like the printing press and the …
IBM has released its Granite 3.2 large language model family featuring a new approach called conditional reasoning. According to Sean Michael Kerner of VentureBeat, this update embeds reasoning capabilities directly into core models rather than creating separate reasoning models. The system allows users to activate reasoning only when needed, improving efficiency for complex tasks. Granite …
Microsoft is now providing unlimited access to Voice and Think Deeper features for all Copilot users at no cost. The Copilot Team announced these advanced capabilities as part of their effort to build “the ultimate AI companion.” Voice allows for extended voice conversations with the AI, while Think Deeper, powered by OpenAI’s o1 model, handles …
Recent developments in AI reasoning capabilities reveal a complex picture of artificial intelligence that excels at certain complex tasks while failing at seemingly simple problems, according to an analysis by Sigal Samuel published in Vox. The article examines claims by leading AI companies about their models’ ability to perform genuine reasoning. OpenAI and other companies …