InfoQ Homepage Large language models Content on InfoQ
-
Google Opens Gemma 4 Under Apache 2.0 with Multimodal and Agentic Capabilities
Google has announced the release of Gemma 4, a series of open-weight AI models, including variants with 2B, 4B, 26B, and 31B parameters, under the Apache 2.0 license. Key features include enhanced video and image processing, audio input on smaller models, and extended context windows up to 256K tokens.
-
Cloudflare Launches Code Mode MCP Server to Optimize Token Usage for AI Agents
Cloudflare has launched a new Model Context Protocol (MCP) server powered by Code Mode, enabling AI agents to interact with large APIs with minimal token usage. The server reduces context footprint across 2,500+ endpoints, improves multi-API orchestration, and provides a secure, code-centric execution environment for LLM agents.
-
Cursor 3 Introduces Agent-First Interface, Moving beyond the IDE Model
Anysphere released Cursor 3, a redesigned interface built from scratch that shifts the primary model from file editing to managing parallel coding agents. The new workspace supports local-to-cloud agent handoff, multi-repo parallel execution, and a plugin marketplace. Community reaction has been divided, with developers questioning cost overhead and the move away from Cursor's IDE-first identity.
-
Google’s TurboQuant Compression May Support Faster Inference, Same Accuracy on Less Capable Hardware
Google Research unveiled TurboQuant, a novel quantization algorithm that compresses large language models’ Key-Value caches by up to 6x. With 3.5-bit compression, near-zero accuracy loss, and no retraining needed, it allows developers to run massive context windows on significantly more modest hardware than previously required. Early community benchmarks confirm significant efficiency gains.
-
Anthropic Paper Examines Behavioral Impact of Emotion-Like Mechanisms in LLMs
A recent paper from Anthropic examines how large language models internally represent concepts related to emotions and how these representations influence behavior. The work is part of the company’s interpretability research and focuses on analyzing internal activations in Claude Sonnet 4.5 to understand the mechanisms behind model responses better.
-
Google Released Gemma 4 with a Focus on Local-First, On-Device AI Inference
With the release of Gemma 4, Google aims to enable local, agentic AI for Android development through a family of models designed to support the entire software lifecycle, from coding to production.
-
Lyft Scales Global Localization Using AI and Human-in-the-Loop Review
Lyft has implemented an AI-driven localization system to accelerate translations of its app and web content. Using a dual-path pipeline with large language models and human review, the system processes most content in minutes, improves international release speed, ensures brand consistency, and handles complex cases like regional idioms and legal messaging efficiently.
-
Google Open Sources Experimental Multi-Agent Orchestration Testbed Scion
Designed to manage concurrent agents running in containers across local and remote compute, Scion is an experimental orchestration testbed that enables developers to run groups of specialized agents with isolated identities, credentials, and shared workspaces.
-
OpenAI Extends the Responses API to Serve as a Foundation for Autonomous Agents
OpenAI announced they are extending the Responses API to make it easier for developer to build agentic workflows, adding support for a shell tool, a built-in agent execution loop, a hosted container workspace, context compaction, and reusable agent skills.
-
Green IT: How to Reduce the Impact of AI on the Environment
AI poses major challenges for green IT: each query consumes vast energy, GPU chips last only 2-3 years, and costs stay hidden from users. Regulatory frameworks like the EU AI Act fall short on enforcement, Ludi Akue said. In her talk What I Wish I Knew When I Started with Green IT she presented model compression, quantization, and novel architectures, using sustainability as a design constraint.
-
QCon London 2026: Ethical AI is an Engineering Problem
At QCon London 2026, Clara Higuera, responsible AI program lead at BBVA, presented how many of the risks associated with AI systems are fundamentally engineering challenges rather than purely governance or policy issues.
-
Apple Improves Context Window Management for its Foundation Models
iOS 26.4, now in Release Candidate, introduces improved context window management for Apple's Foundation Models, helping developers work with the 4096-token context window limit. This encourages treating the context window as a constrained resource, which requires actively managing it like memory in a low-resource system to optimize its usage.
-
Sonatype Launches Guide to Enhance Safety in AI-Assisted Code Generation
Sonatype Guide is a real-time guardrail system that sits between AI coding tools and the open-source ecosystem, ensuring AI-generated code uses safe, valid, and maintainable dependencies.
-
Stripe Engineers Deploy Minions, Autonomous Agents Producing Thousands of Pull Requests Weekly
Stripe engineers describe Minions, autonomous coding agents generating over 1,300 pull requests per week. Tasks can originate from Slack, bug reports, or feature requests. Using LLMs, blueprints, and CI/CD pipelines, Minions produce production-ready changes while maintaining reliability and human review.
-
QCon London 2026: Refreshing Stale Code Intelligence
At QCon London 2026, Jeff Smith discussed the growing mismatch between AI coding models and real-world software development. While AI tools are enabling developers to generate code faster than ever, Smith argued that the models themselves are increasingly “stale” because they lack the repository-specific knowledge required to produce production-ready contributions.