PullRepo

Daily radar for the fastest-growing AI tools & repos

Today's LLM & Language Models: Fastest-Growing Projects — November 22, 2026

Today's trend in the LLM & Language Models space continues to highlight the growing interest in open-source projects that enhance inference capabilities and provide detailed documentation of experiments. Notably, RunanywhereAI/trnrt stands out with its innovative approach to AWS Trainium optimization for large language models (LLMs). Meanwhile, RikiyaT/lm_understand-logs offers a comprehensive resource for those interested in understanding how LLMs can process neural network weights through experimental logs.

RunanywhereAI/trnrt is an open-source project that serves as an inference engine for AWS Trainium, providing NKI megakernels and support for 1M-token context with an OpenAI-compatible API. Its growth score of 49.80 indicates significant interest from the community in optimizing LLMs on specific hardware platforms like AWS Trainium.

RikiyaT/lm_understand-logs provides complete logs detailing experiments that explore how large language models can understand and write neural network weights, contributing to a deeper understanding of model behavior. With over 1,284 stars, this repository has attracted substantial attention for its detailed documentation and insights into the capabilities of LLMs.

pgmergulhao75-dot/openbb-calvona-market-research offers structured market data analysis and research workflows through OpenBB Calvona Market Research, aiming to streamline financial data processing. Despite a lower growth score of 29.25, its increasing number of commits (14 in the last month) suggests active development and community engagement.

affifuddin-hazam/ghost-infer is a fast local LLM inference solution designed for GPUs, featuring Rust binaries, GGUF models, an OpenAI-compatible server, and a chat interface. Its high growth score of 17.54 reflects the growing demand for efficient GPU-based inference solutions that offer compatibility with established APIs.

420420skun-svg/Gemini-Pro-Release presents itself as an advanced AI assistant platform for various tasks including writing, coding, research, and content creation. Despite a lower growth score of 7.70, the project continues to attract interest, indicated by its 65 stars and occasional commits.

zhuangzhaoyeye/llm_course provides a comprehensive course for algorithm engineers focusing on large model mechanisms with practical numerical examples. With a moderate growth score of 7.45, it suggests ongoing development and a growing audience interested in detailed technical education about LLMs.

haydenschmitty/floorplan offers an interactive map of codebases that can visually represent how applications are structured using AI assistance like Claude or Codex. Its impressive star count (5,961) and high number of recent commits (100 in the last month) reflect its popularity as a tool for understanding complex software architectures.

mazulo/leis-mcp serves as an MCP server to facilitate searching and citing Brazilian legislation within AI assistants like Claude. With 71 stars and ongoing development, it demonstrates steady growth in serving specialized legal content through AI interfaces.

DiffuSpace/Clover-35B-A3B-Preview is a diffusion language model with a capacity of 35 billion parameters. Despite a lower growth score (4.60), its small but dedicated community continues to contribute and improve upon the project's offerings.

lixiaoxiao9888-create/forest-laoli-h3 provides a platform-agnostic pipeline for producing short dramas using MiniMax H3, including machine QC processes. With 94 stars, it indicates growing interest in its unique approach to content creation leveraging AI-driven workflows.

These projects collectively illustrate the diverse applications and advancements being made within the LLM & Language Models domain, ranging from hardware optimization and experimental analysis to specialized use cases like market research and legal documentation.
Back to all reports