InfoQ Homepage Large language models Content on InfoQ
-
Hugging Face Introduces mmBERT, a Multilingual Encoder for 1,800+ Languages
Hugging Face has released mmBERT, a new multilingual encoder trained on more than 3 trillion tokens across 1,833 languages. The model builds on the ModernBERT architecture and is the first to significantly improve upon XLM-R, a long-time baseline for multilingual understanding tasks.
-
Google's Agent Development Kit for Java Adds Integration with LangChain4j
The latest release of the Agent Development Kit for Java, version 0.2.0, marks a significant expansion of its capabilities through the integration with the LangChain4j LLM framework, which opens it up to all the large language models supported by the framework.
-
Report Finds LLMs Not Yet Ready to Replace SREs in Incident Management
A study by ClickHouse found that large language models (LLMs) can't yet replace Site Reliability Engineers (SREs) for tasks such as finding the root causes of incidents. The study tested five leading models against real-world observability data to determine whether AI could autonomously identify production issues.
-
xAI Releases Grok 4 Fast with Lower Cost Reasoning Model
xAI has introduced Grok 4 Fast, a new reasoning model designed for efficiency and lower cost.
-
Google Introduces VaultGemma: An Experimental Differentially Private LLM
VaultGemma is a 1B-parameter Gemma 2-based LLM that Google trained from scratch using differential privacy (DP) with the aim of preventing the model from memorizing and later regurgitating training data. While still a research model, VaultGemma could enable applications cases in healthcare, finance, legal, and other regulated sectors.
-
Baidu’s PP-OCRv5 Released on Hugging Face, Outperforming VLMs in OCR Benchmarks
Baidu has released PP-OCRv5 on Hugging Face, a new optical character recognition (OCR) model built to outperform large vision-language models (VLMs) in specialized text recognition tasks. Unlike general-purpose architectures such as Gemini 2.5 Pro, Qwen2.5-VL, or GPT-4o, which handle OCR as part of broader multimodal workflows, PP-OCRv5 is purpose-built for accuracy, efficiency, and speed.
-
Hugging Face Brings Open-Source LLMs to GitHub Copilot Chat in VS Code
Hugging Face has introduced a new integration that allows developers to connect Inference Providers directly with GitHub Copilot Chat in Visual Studio Code. The update means that open-source large language models — including Kimi K2, DeepSeek V3.1, GLM 4.5, and others — can now be accessed and tested from inside the VS Code editor, without the need to switch platforms or juggle multiple tools.
-
Kaggle Introduces Game Arena to Benchmark AI Models in Strategic Games
Kaggle, in collaboration with Google DeepMind, has introduced Kaggle Game Arena, a platform designed to evaluate artificial intelligence models by testing their performance in strategy-based games.
-
Hugging Face Releases FinePDFs: a 3-Trillion-Token Dataset Built from PDFs
Hugging Face has unveiled FinePDFs, the largest publicly available corpus built entirely from PDFs. The dataset spans 475 million documents in 1,733 languages, totaling roughly 3 trillion tokens. At 3.65 terabytes in size, FinePDFs introduces a new dimension to open training datasets by tapping into a resource long considered too complex and expensive to process.
-
Cloudflare Introduces Automated Scoring for Shadow AI Risk Assessment
During AI Week 2025, Cloudflare announced Application Confidence Scores, an automated assessment system that is designed to help organizations evaluate the safety and security of third-party AI applications at scale.
-
Vercel Introduces AI Gateway for Multi-Model Integration
Vercel has rolled out the AI Gateway for production workloads. The service provides a single API endpoint for accessing a wide range of large language and generative models, aiming to simplify integration and management for developers.
-
Google DeepMind Launches EmbeddingGemma, an Open Model for On-Device Embeddings
Google DeepMind has introduced EmbeddingGemma, a 308M parameter open embedding model designed to run efficiently on-device. The model aims to make applications like retrieval-augmented generation (RAG), semantic search, and text classification accessible without the need for a server or internet connection.
-
xAI Releases Grok Code Fast 1, a New Model for Agentic Coding
xAI introduced grok-code-fast-1, a model developed specifically for agentic coding workflows.
-
Google Launches Gemini 2.5 Flash Image with Advanced Editing and Consistency Features
Google released Gemini 2.5 Flash Image (nicknamed nano-banana), its newest image generation and editing model. The system introduces several upgrades over earlier Flash models, including character consistency across prompts, multi-image fusion, precise prompt-based editing, and integration of world knowledge for semantic understanding.
-
LinkedIn Re-Architects Edge-Building System to Support Diverse Inference Workflows
LinkedIn has detailed its re-architected edge-building system, an evolution designed to support diverse inference workflows for delivering fresher and more personalized recommendations to members worldwide. The new architecture addresses growing demands for real-time scalability, cost efficiency, and flexibility across its global platform.