AI Intel

The GLM-5.3-Flash (ox-alpha) model, which reportedly rivals Claude Opus 4.8 on coding and agentic benchmarks, can now be run locally in 3-bit quantiz…

Kimi K3, an open-weight model by Kimi_Moonshot, is now generally available and integrated into GitHub Copilot, demonstrating frontier-level abilities…

Google has rolled out significant updates to Gemini Code Assist, including the General Availability of Gemini 2.5 Pro and Flash models, a preview of…

A new arXiv paper introduces CP-Agent, an AI-powered agent designed to automate and streamline crystal plasticity simulation workflows, which are typ…

A discussion on Hacker News highlights "Jeeves," a system that improves decision models similar to Jev by incorporating advanced reasoning capabiliti…

It appears the era of subsidized compute for ChatGPT Pro is ending, as the $200 plan's compute limits will be halved, and a new $500 plan will offer…

A new arXiv paper introduces "Witeness Overlap" as a method to track directional provenance within open-weight model families, crucial for understand…

A new arXiv paper explores the complex interactions between document parsers, chunking strategies, and embedding models within Retrieval-Augmented Ge…

A new arXiv paper introduces EEGAgentBench, a benchmark designed to evaluate Large Language Model (LLM) agents on both short- and long-horizon electr…

A developer successfully fine-tuned a 4B parameter model named ImaJev-4b over 15 days to excel at making business decisions from text and photos, ach…