Today's LLM & Language Models: Fastest-Growing Projects — May 17, 2026
This week, the landscape of Large Language Models (LLMs) and language models continues to evolve with a mix of innovative projects ranging from novel model architectures to comprehensive datasets and evaluation benchmarks. Among these, several repositories are gaining traction due to their unique approaches and growing community support.
jingyaogong/minimind-o is an ambitious project aimed at training a compact yet versatile 0.1B Omni model capable of multimodal tasks such as listening, speaking, and seeing. With its impressive growth score of 58.91 and 1,321 stars, it stands out for its potential to democratize access to advanced AI capabilities with minimal resources.
ahammadmejbah/Awesome-LLM-Datasets offers a curated collection of datasets tailored for various LLM applications, including medical AI, NLP, multimodal learning, and more. This repository's growth score of 56.00 and 68 stars reflect its value in providing researchers and developers with essential data resources to enhance their model training and evaluation processes.
chiennv2000/orthrus introduces a novel approach for lossless LLM inference through dual-view diffusion decoding, aiming to accelerate the process while preserving accuracy. With a growth score of 38.62 and 219 stars, it demonstrates growing interest in optimizing large model deployment and performance.
alchaincyf/codex-orange-book serves as an extensive guide to OpenAI's Codex technology, offering insights into AI programming practices in the context of GPT-5.5. Its growth score of 36.67 and 136 stars indicate a rising demand for comprehensive resources on integrating advanced language models into software development workflows.
DestinyLinker/MingLi-Bench is a specialized benchmark designed to evaluate LLMs in the context of traditional Chinese fortune-telling methods such as Bazi (八字) and Ziwei Doushu (紫微斗数). With 1,122 stars and a growth score of 34.62, it highlights the expanding scope of AI applications beyond conventional domains.
Continuum-AI-Corp/OrcaRouter-Lite provides a self-hosted solution for managing LLMs with features like streaming support and managed safety nets, making it compatible with OpenAI services while allowing BYOK (Bring Your Own Key) configurations. Its growth score of 29.04 and 351 stars underscore its appeal to users seeking greater control over their AI infrastructure.
shefys/grok-dreamcore is a surreal narrative generator that creates dream sequences based on emotional states, memory anchors, and symbolic archetypes. With a growth score of 27.12 and 311 stars, it showcases the creativity and versatility of language models in generating imaginative content.
rahilp/second-brain-cloudflare offers an innovative layer for integrating various AI tools into a unified memory system accessible through platforms like Claude or ChatGPT. Its growth score of 21.75 and 63 stars reflect its utility in enhancing the efficiency and integration of diverse AI applications, especially on Cloudflare's free tier.
tronghieu/lumina-wiki aims to be an AI-powered research assistant capable of organizing and connecting knowledge across various domains. With a growth score of 20.84 and 28 stars, it highlights the growing interest in leveraging language models for advanced knowledge management and retrieval systems.
Beever-AI/beever-atlas is designed as an LLM-based conversation knowledge base, facilitating efficient information recall and organization. Its growth score of 20.15 and 328 stars indicate its potential to streamline interactions with large language models in a user-friendly manner, making it accessible for broader audiences.
These projects collectively illustrate the dynamic nature of the LLM ecosystem, reflecting ongoing innovations in model architecture, dataset curation, inference optimization, and application-specific benchmarks.
jingyaogong/minimind-o is an ambitious project aimed at training a compact yet versatile 0.1B Omni model capable of multimodal tasks such as listening, speaking, and seeing. With its impressive growth score of 58.91 and 1,321 stars, it stands out for its potential to democratize access to advanced AI capabilities with minimal resources.
ahammadmejbah/Awesome-LLM-Datasets offers a curated collection of datasets tailored for various LLM applications, including medical AI, NLP, multimodal learning, and more. This repository's growth score of 56.00 and 68 stars reflect its value in providing researchers and developers with essential data resources to enhance their model training and evaluation processes.
chiennv2000/orthrus introduces a novel approach for lossless LLM inference through dual-view diffusion decoding, aiming to accelerate the process while preserving accuracy. With a growth score of 38.62 and 219 stars, it demonstrates growing interest in optimizing large model deployment and performance.
alchaincyf/codex-orange-book serves as an extensive guide to OpenAI's Codex technology, offering insights into AI programming practices in the context of GPT-5.5. Its growth score of 36.67 and 136 stars indicate a rising demand for comprehensive resources on integrating advanced language models into software development workflows.
DestinyLinker/MingLi-Bench is a specialized benchmark designed to evaluate LLMs in the context of traditional Chinese fortune-telling methods such as Bazi (八字) and Ziwei Doushu (紫微斗数). With 1,122 stars and a growth score of 34.62, it highlights the expanding scope of AI applications beyond conventional domains.
Continuum-AI-Corp/OrcaRouter-Lite provides a self-hosted solution for managing LLMs with features like streaming support and managed safety nets, making it compatible with OpenAI services while allowing BYOK (Bring Your Own Key) configurations. Its growth score of 29.04 and 351 stars underscore its appeal to users seeking greater control over their AI infrastructure.
shefys/grok-dreamcore is a surreal narrative generator that creates dream sequences based on emotional states, memory anchors, and symbolic archetypes. With a growth score of 27.12 and 311 stars, it showcases the creativity and versatility of language models in generating imaginative content.
rahilp/second-brain-cloudflare offers an innovative layer for integrating various AI tools into a unified memory system accessible through platforms like Claude or ChatGPT. Its growth score of 21.75 and 63 stars reflect its utility in enhancing the efficiency and integration of diverse AI applications, especially on Cloudflare's free tier.
tronghieu/lumina-wiki aims to be an AI-powered research assistant capable of organizing and connecting knowledge across various domains. With a growth score of 20.84 and 28 stars, it highlights the growing interest in leveraging language models for advanced knowledge management and retrieval systems.
Beever-AI/beever-atlas is designed as an LLM-based conversation knowledge base, facilitating efficient information recall and organization. Its growth score of 20.15 and 328 stars indicate its potential to streamline interactions with large language models in a user-friendly manner, making it accessible for broader audiences.
These projects collectively illustrate the dynamic nature of the LLM ecosystem, reflecting ongoing innovations in model architecture, dataset curation, inference optimization, and application-specific benchmarks.