LLM Pulse- March 06, 2025

New Releases & Updates

OpenAI launching GPT-4.5, its next general-purpose large language model: OpenAI released a new base model on Thursday called GPT-4.5, which the company said is its best and smartest model for chat yet. It’s not a reasoning model like OpenAI’s o1 and o3 models, but it can be used to train other models to be reasoning models. Notably, GPT-4.5 was trained using 10 times the computing power (scores of GPUs in data centers) than its predecessor, GPT-4o. Read More

Grok 3 Released - NewsElon Musk’s xAI launches Grok-3: What you need to know: Elon Musk’s artificial intelligence startup, xAI, has launched the most powerful iteration of its Large Language Model (LLM), Grok-3. Grok-3’s was formally announced in a livestream on social media platform X. Read More

Gemini Advanced explained: Availability, language support, and more: Google’s Gemini brand has taken on many personas since its debut as only an AI language model. Today, it refers to a chatbot of the same name, a whole family of language models, various Android features, and a monthly subscription. Read More

Staqu Unveils Jarvis GPT: A Powerful Fusion of Vision AI and LLM for Smart Retail Analytics: Combining advanced video analytics with the power of large language models, Jarvis GPT delivers real-time insights, optimizing retail operations and enhancing customer experiences. Read More

Fetch.ai Launches First Web3? LLM? ‘ASI-1 Mini’ for Agentic AI: The race to master the onchain vertical filled by AI agents is heating up with a familiar player having just thrown down the gauntlet. Fetch.ai, one of the best known and longest-standing blockchain projects developing artificial intelligence solutions, has unveiled ASI-1 Mini – and it has the potential to do for web3 what ChatGPT’s o3-mini has done for centralized AI. Read More

Permutable AI Launches LLM-Driven Market 360 Thematics for Cross-Asset Sentiment Analysis: Share this post:Permutable AI, a specialist in LLM-driven market sentiment analysis today announced the launch of Market 360, a market sentiment visualization tool that provides a comprehensive 360-degree view across commodity. Read More

?GSMA launches Open-Telco LLM Benchmarks: The GSMA Foundry has launched GSMA Open-Telco LLM Benchmarks, an open-source community aimed at improving the performance of large language models (LLMs) for telecom-specific applications. Read More

?Mistral's new AI model specializes in Arabic and related languages Saba is just the latest region-specific LLM: Paris-based AI startup Mistral is focusing on providing large language models (LLMs) that understand regional-specific languages and are tailored to grasp the cultural nuances sometimes overlooked in larger, more general-purpose models. Read More

It is 10 times faster than GPT-4o: Inception Labs unveils Mercury — the first diffusion speech model: For a long time, there have been active discussions about finding a better architecture for large language models (LLMs) that could become an alternative to transformers. It seems that California-based startup Inception Labs already has a promising solution. The company has introduced Mercury, the world’s first diffusion-based large language model designed for commercial use. Read More

LangWatch bags €1 million for reliable and fast AI Optimization platform: Amsterdam-based LangWatch, a startup building the “world’s first” LLMops (Large Language Model Operations) platform to monitor, evaluate and optimise LLM-powered applications, has raised a €1 million pre-Seed round. Read More

Research and Technology

What Is a Diffusion LLM and Why Does It Matter?: What is diffusion large language model LLM, and why it matters. In the context of Inception Labs releasing Mercury Coder. Maybe an alternative to the leading auto-regression LLMs (chatGPT, Claude, Gemini). Read More

Like human brains, large language models reason about diverse data in a general way: A new study shows LLMs represent different data types based on their underlying meaning and reason about data in their dominant language. Read More

Beyond ChatGPT’s Extension: How to Redirect Safari Searches to Any LLM: Earlier this week, OpenAI’s official ChatGPT app for iPhone and iPad was updated with a native Safari extension that lets you forward any search query from Safari’s address bar to ChatGPT Search. It’s a clever approach. Read More

LLM Mistral rivals GPT-4 Turbo for extracting clinical history elements: Large language model (LLM) Mistral outperformed Llama and GPT-4 Turbo in a real-world application that assessed the completeness of clinical histories accompanying radiology imaging orders from the emergency department. Read More

LLM continuous self-instruct fine-tuning framework powered by a compound AI system on Amazon SageMaker: Fine-tuning a pre-trained large language model (LLM) allows users to customize the model to perform better on domain-specific tasks or align more closely with human preferences. It is a continuous process to keep the fine-tuned model accurate and effective in changing environments. Read More

LLM-based web application scanner recognizes tasks and workflows: A new automated web application scanner autonomously understands and executes tasks and workflows on web applications. The tool named YuraScanner harnesses the world knowledge stored in large language models (LLMs) to navigate through web applications in the same way a human user would. Read More

The ‘first commercial scale’ diffusion LLM mercury offers over 1000 tokens/sec on nvidia h100: For a long time, there’s been an active discussion about exploring a better architecture for large language models (LLM) besides the transformer. Well, two months into 2025, this California-based startup seems to have a promising solution.? Read More

Comparing the performance of a large language model and naive human interviewers in interviewing children about a witnessed mock-event: The present study compared the performance of a Large Language Model (LLM; ChatGPT) and human interviewers in interviewing children about a mock-event they witnessed. Read More

MicroCloud Hologram Inc. Achieves Breakthrough in Optimizing Scaling Methods for Open-Source Configurations Using Deepseek LLM: MicroCloud Hologram Inc. (NASDAQ: HOLO), ("HOLO" or the "Company"), a technology service provider, delved deeply into scaling laws and made unique discoveries, providing key support for the scaling of large models in two commonly used open-source configurations: 7B and 67B. Read More

Curious about DeepSeek but worried about privacy? These apps let you use an? LLM? without the internet: Most of us are used to using internet chatbots like ChatGPT and DeepSeek in one of two ways: via a web browser or via their dedicated smartphone apps. There are two drawbacks to this. First, their use requires an internet connection. Second, everything you type into the chatbot is sent to the companies’ servers, where it is analyzed and retained. In other words: the more you use the chatbot the more the company knows about you. Read More

AI Agents: Why Workflows Are the LLM Use Case to Watch: There’s an unsung hero of the modern enterprise: the workflow. It’s sometimes called a rules engine, a process flow, a single-state machine or a software-defined workflow. In a user interface (UI), it’s a “wizard.” Developers often call it. Read More

Explained: The? LLM? design strategy that helped Deep seek rock the AI world: The sudden rise of Chinese AI start-up DeepSeek has taken the AI industry by surprise. Despite being just two years old, the company's large language models (LLMs) are on par with those of AI giants like OpenAI, Google DeepMind, xAI, and others. DeepSeek’s R-1 and V-3 models have outperformed OpenAI’s GPT-4o and O3 Preview, Google’s Gemini Pro Flash, and Anthropic’s Claude 3.5 Sonnet across various benchmarks. Read More

6 common LLM customization strategies briefly explained: Training a LLM model from scratch is largely infeasible to small to medium teams due to the demand of massive amounts of training data and resources. Therefore, a wide range of LLM customization strategies are developed in recent years to tune the models for various scenarios that require specialized knowledge. Read More

Other News?

Large Language Models pose growing security risks: More powerful and pervasive large language models are creating a new cybersecurity challenge for companies. The risks posed by LLMs, a form of generative artificial intelligence that communicates through language in a humanlike way, are already manifold. Read More

Large Language Model (llm) market is set to fly high growth in years to come: According to USD Analytics the Global Large Language Model (LLM) Market is expected to reach 66.8 Billion USD by 2034 from 19 Billion USD in 2025, at a CAGR of 17.02%. Read More

12K hardcoded API keys and passwords found in public LLM training data: Roughly 12,000 hardcoded live API keys and passwords were found on Common Crawl, a large dataset used to train LLMs such as DeepSeek. Security pros say hardcoded credentials are dangerous because hackers can more easily exploit them to gain access to sensitive data, systems, and networks. Read More

'More Than Words' Review: When AI Is the Author: Writing cannot be taught but it can be learned. A paradox? A Zen koan? Perhaps both, but this is what I concluded after 30 years of teaching a course called Advanced Prose Composition at Northwestern University. I am not about to say that I was a good teacher, for one of the few truths about teaching I know is that those who think they teach well probably don't. Read More

Korea hopes gov’t-led? LLM? will reposition it in AI landscape: The South Korean government announced plans to mobilize national resources in a bid to develop a large-scale language model (LLM) comparable to OpenAI’s ChatGPT. The initiative aims to concentrate the country’s resources and talent in artificial intelligence (AI) to catch up with leading global players. If successful, it would mark a significant achievement, particularly considering that China’s DeepSeek has demonstrated that big tech-led AI competition is not insurmountable. But the crucial factor for success lies in the execution of this plan. Read More

Adsure Services announces software contract to enhance AI LLM training and provide efficiencies: Adsure Services has announced that its operating subsidiary, TIAA Ltd, has entered into an agreement with K10 Vision to implement advanced audit working paper software, marking a substantial step forward in Adsure’s digital transformation.? Read More

AKOOL Unveils Enhanced Streaming Avatars with Seamless? LLM? Integration to Revolutionize Human-Machine Interactions: AKOOL Unveils Enhanced Streaming Avatars with Seamless LLM Integration to Revolutionize Human-Machine Interactions. AKOOL, a trailblazer in AI-driven content creation, today announced enhancements to AKOOL Streaming Avatars, an advanced video generation technology that now seamlessly integrates with large language models (LLM) to help model builders create dynamic, lifelike avatars. Read More