InfoQ Homepage Large language models Content on InfoQ
-
Code Arena Launches as a New Benchmark for Real-World AI Coding Performance
LMArena has launched Code Arena, a new evaluation platform that measures AI models' performance in building complete applications instead of just generating code snippets. It emphasizes agentic behavior, allowing models to plan, scaffold, iterate, and refine code within controlled environments that replicate actual development workflows.
-
Anthropic Adds Sandboxing and Web Access to Claude Code for Safer AI-Powered Coding
Anthropic released sandboxing capabilities for Claude Code and launched a web-based version of the tool that runs in isolated cloud environments. The company introduced these features to address security risks that arise when Claude Code writes, tests, and debugs code with broad access to developer codebases and files.
-
Google Unveils Project Suncatcher, Envisioning AI Models Running in Space
Google has unveiled Project Suncatcher, a research initiative exploring how solar powered satellite constellations equipped with Tensor Processing Units TPUs could one day enable large scale artificial intelligence computation in space.
-
New Claude Haiku 4.5 Model Promises Faster Performance at One-Third the Cost
Anthropic released Claude Haiku 4.5, making the model available to all users as its latest entry in the small, fast model category. The company positions the new model as delivering performance levels comparable to Claude Sonnet 4, which launched five months ago as a state-of-the-art model, but at "one-third the cost and more than twice the speed."
-
Anthropic Finds LLMs Can Be Poisoned Using Small Number of Documents
Anthropic's Alignment Science team released a study on poisoning attacks on LLM training. The experiments covered a range of model sizes and datasets, and found that only 250 malicious examples in pre-training data were needed to create a "backdoor" vulnerability. Anthropic concludes that these attacks actually become easier as models scale up.
-
CodeClash Benchmarks LLMs through Multi-Round Coding Competitions
Researchers from Standford, Princeton, and Cornell have developed a new benchmark to better evaluate coding abilities of large language models (LLMs). Called CodeClash, the new benchmark pits LLMs against each other in multi-round tournaments to assess their capacity to achieve competitive, high-level objectives beyond narrowly defined, task-specific problems.
-
GitHub Expands Copilot Ecosystem with AgentHQ
GitHub has announced AgentHQ, a new addition to its platform that aims to unify the fragmented landscape of AI tools within the software development process.
-
Android GenAI Prompt API Enables Natural Language Requests with Gemini Nano
The ML Kit GenAI Prompt API, now available in alpha, enables Android developers to send natural language and multimodal requests to Gemini Nano running on-device, extending the text summarization and image description capabilities introduced with the initial GenAI release.
-
Cursor 2.0 Expands Composer Capabilities for Context-Aware Development
Cursor has launched version 2.0 of its AI-driven code editor, featuring Composer, a new model that enables developers to write and modify code through natural language interaction.
-
Apple Releases Pico-Banana-400K Dataset to Advance Text-Guided Image Editing
Pico-Banana-400K is a curated dataset of 400,000 images developed by Apple researchers to make it easier to create text-guided image editing models. The images were generated using Google's Nano-Banana to modify real photographs from the Open Images collecion and were then filtered using Gemini-2.5-Pro based on their overall quality and prompt compliance.
-
Inside the Architectures Powering Modern AI Systems: QCon San Francisco 2025
Senior engineers face fast-moving AI adoption without clear patterns. QCon SF 2025 brings real-world lessons from teams at Netflix, Meta, Intuit, Anthropic & more, showing how to build reliable AI systems at scale. Early bird ends Nov 11.
-
The Architectural Shift: AI Agents Become Execution Engines While Backends Retreat to Governance
A fundamental shift in enterprise software architecture is emerging as AI agents transition from assistive tools to operational execution engines, with traditional application backends retreating to governance and permission management roles. This transformation is accelerating across sectors, with 40% of enterprise applications expected to include autonomous agents by 2026.
-
NVIDIA Introduces OmniVinci, a Research-Only LLM for Cross-Modal Understanding
NVIDIA has introduced OmniVinci, a large language model designed to understand and reason across multiple input types — including text, vision, audio, and even robotics data. The project, developed by NVIDIA Research, aims to push machine intelligence closer to human-like perception by unifying how models interpret the world across different sensory streams.
-
Anthropic Introduces Skills for Custom Claude Tasks
Anthropic has unveiled a new feature called Skills, designed to let developers extend Claude with modular, reusable task components.
-
DeepSeek AI Unveils DeepSeek-OCR: Vision-Based Context Compression Redefines Long-Text Processing
DeepSeek AI has developed DeepSeek-OCR, an open-source system that uses optical 2D mapping to compress long text passages. This approach aims to improve how large language models (LLMs) handle text-heavy inputs.