AI Intel

The founder of GLM reportedly indicated that the GLM-fable model might be released before the end of the year.

Unsloth has released a GLM-5.2-GGUF model, including a 2-bit quantized version, with a file size of 238GB.

SproutRAG is a new method for long-document Retrieval-Augmented Generation (RAG) that uses attention-guided tree search and progressive embeddings to…

CODEBLOCK is a new approach that addresses the challenge of supervised fine-tuning for code LLMs by learning to apply supervision at the right granul…

Jesse is a new tool that allows users to search the live internet, aiming to replace traditional lead list building methods like Apollo/Clay.

D-ID has launched "Agentic videos," which are interactive videos capable of engaging in conversational AI.

Ollama's v0.30.9 update introduces support for the Cohere2Moe architecture and resolves an issue where coding agents and assistants would not output…

OpenAI Codex v0.141.0 introduces authenticated, end-to-end encrypted Noise relay channels for remote executors and preserves native working directori…

Claude Code v2.1.178 introduces new Tool(param:value) syntax for granular permission rules and enables loading skills from nested .claude/skills dire…

vLLM has released version 0.23.0, a substantial update featuring 408 commits from 200 contributors, though Minimax M3 support is not yet included.