Mistral Releases an Open Source Model that Outperforms Gemma 3 and GPT-4o Mini

Paris–based Mistral AI unveiled Mistral Small 3.1, a new multimodal open-source model. According to the company, it is "the best model in its weight class " and "outperforms comparable models like Gemma 3 and GPT-4o Mini. " Released under an Apache 2.0 license, Mistral Small 3.1 has an expanded context window of up to 128k tokens and a delivery inference speed of 150 tokens per second. Experts say that Mistral Small 3 is competitive with larger models such as Llama 3.3 70B or Qwen 32B and replaces opaque proprietary models like GPT4o-mini. Mistral Small 3 can be fine-tuned to specialize in specific domains, creating highly accurate experts. This is particularly useful in fields like legal advice, medical diagnostics, and technical support, where domain-specific knowledge is essential. This model sets the stage for increased competition in a market dominated by U.S. tech giants. Mistral's open-source approach highlights a growing divide in the AI industry between closed, proprietary systems and open, accessible alternatives. After raising $1.04 billion, founded in 2023 by former researchers from Google DeepMind and Meta, Mistral AI has rapidly established itself as Europe’s leading AI startup, with a valuation of approximately $6 billion. While impressive for a European startup, this valuation remains a fraction of OpenAI’s reported $80 billion. Mistral Small 3.1 joins the company's rapidly expanding suite of AI products. Earlier this month, the company introduced Mistral OCR, an optical character recognition API that converts PDF documents into AI-ready Markdown files. This addresses a critical need for enterprises seeking to make document repositories accessible to AI systems. These specialized tools complement Mistral’s broader portfolio, which includes Mistral Large 2 (their flagship large language model), Pixtral (for multimodal applications), Codestral (for code generation), and “Les Ministraux,” a family of models optimized for edge devices.

OpenAI Integrates New Image Generation Capabililty Into ChatGPT

Google Gemini's Latest Feature: Canvas, an Interactive Space for Refining Documents and Code

DeepSeek Releases a New Iteration of Its Model, MIT-Licensed

The New Trend of 'Vibecoding': Non-Programmers Creating Software Tools with AI

Nontechnical hobbyists without code knowledge are starting to build fully functional apps and websites by typing prompts into AI-driven text boxes in a trend called vibecoding.The term vibecoding has been popularized by the AI researcher Andrej Karpathy, as he explained in this post. There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper… — Andrej Karpathy (@karpathy) February 2, 2025 An expert explained in The New York Times that these aren’t the kinds of tools that a big tech company would build. "There’s no real market for them; their features are limited, and some of them only sort of work." These creations range from tools that transcribe and summarize long podcasts to searchable web databases or practical organization apps. "Building software this way produces a feeling of AI vertigo, similar to what I felt after using ChatGPT for the first time," wrote technologist reporter Kevin Roose. document.createElement('video'); https://iblnews.org/wp-content/uploads/2025/03/134734_1_27roose-video-2_wg_1080p.mp4 In the last year, new tools, such as Cursor, Replit, Bolt, and Lovable, have been built to use more powerful AI models, enabling neophytes to program like pros, as shown in the video above. After the user’s prompt, the tool creates a design, decides on the best software packages and programming languages, and builds a working prototype. Users can suggest tweaks and revisions and then deploy their new product to the web or run it on their computers. Depending on the project's complexity, the process takes a few minutes or several hours. Most of the products allow limited free use, with paid tiers that unlock better features and the ability to build more things. Not all vibecoding experiments have been successful, and occasionally, AI makes mistakes. For now, these tools are for hobby projects or even junior programmers, but not essential tasks. the 14 most valuable skills in startups today: it's crazy that university doesn't teach this: 1. attention arbitrage – spot which platforms are undercharging for reach and get the right creative in front of them 2. thumb-stopping content – understand how to make people stop… — GREG ISENBERG (@gregisenberg) February 18, 2025

Nvidia Presented an Open-Source Humanoid Robot Foundation Model, 'Isaac GR00T N1'

Nvidia presented an open-source humanoid robot called Isaac GR00T N1 at its annual conference yesterday. The company said it was "the world’s first open, fully customizable foundation model for generalized humanoid reasoning and skills." "The age of generalist robotics is here," said Jensen Huang, founder and CEO of NVIDIA. "With Isaac GR00T N1 and new data-generation and robot-learning frameworks, robotics developers everywhere will open the next frontier in the age of AI." Newton, an open-source physics engine — under development with Google DeepMind and Disney Research — purpose-built for developing robots. The GR00T N1 foundation model features a dual-system architecture inspired by human cognition. System 1 is a fast-thinking action model that mirrors human reflexes or intuition, while System 2 is a slow-thinking model for deliberate, methodical decision-making. The chip company said GR00T N1 could generalize everyday tasks — such as grasping, moving objects with one or both arms, and transferring items from one arm to another — or perform multistep tasks requiring extended context and combinations of general skills. These capabilities can be applied to material handling, packaging, and inspection cases. Developers and researchers can post-train GR00T N1 with real or synthetic data for their specific humanoid robot or task. In his GTC keynote, Huang demonstrated 1X’s humanoid robot autonomously performing domestic tidying tasks using a post-trained policy built on GR00T N1. The robot’s autonomous capabilities are the result of an AI training collaboration between 1X and NVIDIA. Among the additional leading humanoid developers worldwide with early access to GR00T N1 are Agility Robotics, Boston Dynamics, Mentee Robotics, and NEURA Robotics. NVIDIA announced GR00T N1, the first fully customizable open-source humanoid robot foundation model, designed to advance general-purpose robotics. ⦿ A dual-system AI inspired by human cognition: System-1 handles fast, intuitive actions, while System-2 enables methodical… pic.twitter.com/qb3gVlMv1o — The Humanoid Hub (@TheHumanoidHub) March 18, 2025

Microsoft Announces Two Sales AI Agents In Competition with Salesforce

Columbia University Conceded to Trump Administration Demands

Trump Signed Order Aimed at Shutting Down the U.S. Department of Education

The 1EdTech Consortium Released a New Comprehensive Learning Record (CLR) for Digital Credentials

The 1EdTech Consortium (1EdTech) released the CLR Standard 2.0, its latest innovation in digital credential technology, during the organization’s Digital Credentials Summit, which took place this month in Phoenix, Arizona. "With CLR Standard 2.0, we are making credentials more secure, portable, and universally verifiable, we are ensuring that every learner has the ability to own and share their achievements with confidence,” said Curtiss Barnes, CEO of 1EdTech. The CLR Standard establishes a common way to package credential records in formal and informal learning, regardless of who provides them to the learner. CLR 2.0 improves portability and verifiability, giving the earner control of their credentials by using rich human-readable and machine-verifiable metadata. CLR 2.0 is also designed as Verifiable Credentials (VCs), as defined by the World Wide Web Consortium (W3C), and aligns with 1EdTech’s Open Badges 3.0 standard, making it even easier to share credentials between wallets. According to a new 1EdTech report, “Six Steps for a Successful Credentialing Program: A Case Study Review,” credentialing platforms that adhere to open standards are one of the six steps successful credentialing programs worldwide recommend.

Experts Say that The Future of U.S. Research Universities Is at Stake

Public and private universities have developed a critical portion of the nation’s research and scholarship in the last 75 years since the National Science Foundation (NSF) launch in 1950. Faculty-led researchers and students have spread doctoral education and become innovators and leaders of society and the world’s largest economy. Of the over 2,000 research universities in the country, 146 are Carnegie R1 (doctoral universities with very high research intensity). Universities collectively spent $97 billion on research, of which $54 billion was federal support. Twenty most prominent universities receive a third of the support. But will the American research universities make it to their 100th birthday? According to Dr Robert A Brown, President Emeritus and Computing and Data Sciences Professor at Boston University, various forces threaten the viability of U.S. research universities. Among those forces, experts note the new political landscape, with the Trump administration's announcements to cut research funds, the decline of college-age students, massive pedagogical and operational upheavals caused by AI, and the expectation that the cost of education will come down with chatbots and copilots replacing staff. Professor Bryan Alexander assembled evidence of the adverse effects felt in the academic world after President Trump took office. "Most of the Institute for Education Sciences (IES) unit, a research team that’s part of the National Center for Education Statistics (NCES), have been laid off. Nearly $1 billion in IES research contracts has been cut. More than $330 million was cut from the Center’s Regional Educational Laboratories and Equity Assistance Centers. At least thousands of students have seen internships with the federal government vanish, or at least become unclear if they will happen, due to cuts." Among other cuts, Bryan Alexander noted these: • MIT, expecting federal losses of $100 million or more, suspended non-faculty hiring. • The University of Louisville has ordered a hiring freeze, apparently across the board, until summer. • North Carolina State University is going to implement a total hiring freeze. • Columbia University’s medical school announced a hiring freeze. • Northwestern University announced a 10% cut to non-personnel expenses. The new administration cut $600 million in DEI-related campus grants.

......

Today's Summary

Sunday, November 23, 2025

Education technology today is marked by rising AI adoption among educators and innovative personalized learning approaches.

Video News

Loading videos...

Today in AI & EdTech

Sunday, November 23, 2025

AI is transforming the education technology landscape as more teachers adopt intelligent tools, driving forward and adaptive learning experiences.

AI & EdTech Videos

OpenAI Launches Educational GPT Model

Adaptive Learning Platforms Show 40% Improvement

Microsoft Education Copilot Beta Launch

Today in Education

U.S. Department of Education Announces New Funding for STEM Programs

The initiative aims to support science, technology, engineering, and mathematics education.

Global Education Summit Highlights Digital Learning Innovations

Leaders from around the world discuss the future of remote and hybrid learning models.

New Study Shows Benefits of Early Childhood Education

Research indicates significant long-term academic and social advantages for students.