Today's LLM & Language Models: Fastest-Growing Projects — April 11, 2026
Today's the LLM & Language Models space, we're seeing a surge of interest in tools that enable on-device processing and compilation of large language models. Developers are eager to tap into the potential of LLMs without relying on cloud services or API keys, driving growth in repositories focused on local inference and knowledge base construction.
Pratiyush's llm-wiki takes the top spot with a Growth Score of 80.83 and 56 stars, offering a Karpathy-inspired LLM-powered knowledge base that can be constructed from various code sessions. Its popularity stems from its ability to provide a structured and searchable wiki, making it an attractive solution for developers looking to organize their knowledge.
Kessler's gemma-gem comes in second with a Growth Score of 77.92 and 638 stars, showcasing Gemma 4 model processing entirely on-device via WebGPU. This innovative approach eliminates the need for API keys or cloud services, resonating with users seeking more control over their data.
The sdyckjq-lab's llm-wiki-skill repository boasts a Growth Score of 70.83 and 511 stars, providing a personal knowledge base construction skill based on Karpathy's method. Its support for multiple platforms has likely contributed to its growing popularity among developers seeking flexible solutions.
Xoai's sage-wiki, with a Growth Score of 68.07 and 356 stars, offers an LLM-compiled personal knowledge base that compiles papers, articles, and notes into a structured wiki. The repository's growth can be attributed to its ability to extract concepts, discover cross-references, and provide search functionality.
Mnfst's awesome-free-llm-apis takes the fifth spot with a Growth Score of 65.98 and an impressive 2,096 stars, providing a comprehensive list of permanent free LLM APIs. The repository's popularity stems from its value as a resource for developers seeking cost-effective solutions.
Quantumaikr's quant.cpp, with a Growth Score of 39.36 and 373 stars, offers high-performance LLM inference in pure C with zero dependencies. Its innovative approach to lossless KV cache compression has likely driven interest among developers seeking efficient solutions.
M0at's rvllm repository boasts a Growth Score of 38.37 and 416 stars, providing high-performance LLM inference in Rust as a drop-in replacement for vLLM. The growth of this repository can be attributed to its potential for seamless integration into existing projects.
Zolotukhin's zinc, with a Growth Score of 27.24 and 299 stars, offers local LLM inference on AMD GPUs and Apple Silicon using Zig. Although it has a lower growth score compared to other repositories, its unique approach has likely piqued the interest of developers exploring alternative solutions.
Rasbt's llm-architecture-gallery repository has a Growth Score of 23.94 and 1,020 stars, providing source data for LLM architecture visualization. Its popularity can be attributed to the growing interest in understanding LLM architectures and their applications.
Kreuzberg-dev's liter-llm rounds out the list with a Growth Score of 22.50 and 141 stars, offering a universal LLM API client with support for over 142 providers and 11 native language bindings powered by Rust core. Its growth is likely driven by its potential to simplify interactions with multiple LLM APIs.
Pratiyush's llm-wiki takes the top spot with a Growth Score of 80.83 and 56 stars, offering a Karpathy-inspired LLM-powered knowledge base that can be constructed from various code sessions. Its popularity stems from its ability to provide a structured and searchable wiki, making it an attractive solution for developers looking to organize their knowledge.
Kessler's gemma-gem comes in second with a Growth Score of 77.92 and 638 stars, showcasing Gemma 4 model processing entirely on-device via WebGPU. This innovative approach eliminates the need for API keys or cloud services, resonating with users seeking more control over their data.
The sdyckjq-lab's llm-wiki-skill repository boasts a Growth Score of 70.83 and 511 stars, providing a personal knowledge base construction skill based on Karpathy's method. Its support for multiple platforms has likely contributed to its growing popularity among developers seeking flexible solutions.
Xoai's sage-wiki, with a Growth Score of 68.07 and 356 stars, offers an LLM-compiled personal knowledge base that compiles papers, articles, and notes into a structured wiki. The repository's growth can be attributed to its ability to extract concepts, discover cross-references, and provide search functionality.
Mnfst's awesome-free-llm-apis takes the fifth spot with a Growth Score of 65.98 and an impressive 2,096 stars, providing a comprehensive list of permanent free LLM APIs. The repository's popularity stems from its value as a resource for developers seeking cost-effective solutions.
Quantumaikr's quant.cpp, with a Growth Score of 39.36 and 373 stars, offers high-performance LLM inference in pure C with zero dependencies. Its innovative approach to lossless KV cache compression has likely driven interest among developers seeking efficient solutions.
M0at's rvllm repository boasts a Growth Score of 38.37 and 416 stars, providing high-performance LLM inference in Rust as a drop-in replacement for vLLM. The growth of this repository can be attributed to its potential for seamless integration into existing projects.
Zolotukhin's zinc, with a Growth Score of 27.24 and 299 stars, offers local LLM inference on AMD GPUs and Apple Silicon using Zig. Although it has a lower growth score compared to other repositories, its unique approach has likely piqued the interest of developers exploring alternative solutions.
Rasbt's llm-architecture-gallery repository has a Growth Score of 23.94 and 1,020 stars, providing source data for LLM architecture visualization. Its popularity can be attributed to the growing interest in understanding LLM architectures and their applications.
Kreuzberg-dev's liter-llm rounds out the list with a Growth Score of 22.50 and 141 stars, offering a universal LLM API client with support for over 142 providers and 11 native language bindings powered by Rust core. Its growth is likely driven by its potential to simplify interactions with multiple LLM APIs.