Today's LLM & Language Models: Fastest-Growing Projects — April 21, 2026
Today's the LLM & Language Models space, we're seeing a surge in repositories focused on building personal knowledge bases and making language models more conversational. The trend is clear: developers are eager to harness the power of LLMs to organize their thoughts, automate tedious tasks, and create more human-like interactions. With growth scores soaring across the board, it's an exciting time for innovation in this space.
arman-bd/guppylm, with a staggering growth score of 83.63 and over 2,962 stars, is making waves as a ~9M parameter LLM that talks like a small fish. Its unique approach to language generation has captured the attention of developers looking to experiment with more whimsical models.
hexiecs/talk-normal, boasting a growth score of 68.92 and 1,345 stars, aims to make any LLM talk like a normal person by removing AI slop. As developers seek to refine their models' communication skills, this repository is gaining traction as a go-to solution for achieving more natural-sounding interactions.
sdyckjq-lab/llm-wiki-skill, with a growth score of 54.84 and 921 stars, offers a personal knowledge base construction skill based on Karpathy's llm-wiki methodology. Its support for multiple platforms has resonated with developers seeking to centralize their knowledge management.
amitshekhariitbhu/llm-internals, sporting a growth score of 43.83 and 588 stars, provides an in-depth exploration of LLM internals, from tokenization to attention to inference optimization. As interest in understanding the inner workings of these complex models grows, this repository is attracting developers eager to learn.
kessler/gemma-gem, with a growth score of 37.22 and 801 stars, enables users to run Google's Gemma 4 model entirely on-device via WebGPU — eliminating the need for API keys, cloud services, or data transmission. Its emphasis on security and self-sufficiency has drawn attention from developers prioritizing data protection.
xoai/sage-wiki, boasting a growth score of 36.97 and 438 stars, compiles papers, articles, and notes into a structured, interlinked wiki using an LLM. Its innovative approach to knowledge organization is resonating with researchers and developers seeking to streamline their information management.
VectifyAI/OpenKB, with a growth score of 32.76 and 445 stars, presents itself as an open LLM knowledge base, offering a centralized hub for storing and retrieving information. As the demand for transparent and accessible knowledge repositories grows, this repository is gaining traction.
lucasastorian/llmwiki, sporting a growth score of 31.29 and 575 stars, provides an open-source implementation of Karpathy's LLM Wiki pattern, allowing users to upload documents and connect their Claude account via MCP. Its flexibility and adaptability have earned it a loyal following among developers.
Pratiyush/llm-wiki, with a growth score of 30.12 and 132 stars, implements the LLM-powered knowledge base concept using Claude Code, Codex CLI, Copilot, Cursor & Gemini sessions. Although smaller in size, this repository is growing steadily as developers explore its potential for centralized information management.
atomicmemory/llm-wiki-compiler, boasting a growth score of 29.72 and 633 stars, compiles raw sources into an interlinked wiki using an LLM, inspired by Karpathy's pattern. Its straightforward approach to knowledge organization has attracted attention from developers seeking efficient solutions.
arman-bd/guppylm, with a staggering growth score of 83.63 and over 2,962 stars, is making waves as a ~9M parameter LLM that talks like a small fish. Its unique approach to language generation has captured the attention of developers looking to experiment with more whimsical models.
hexiecs/talk-normal, boasting a growth score of 68.92 and 1,345 stars, aims to make any LLM talk like a normal person by removing AI slop. As developers seek to refine their models' communication skills, this repository is gaining traction as a go-to solution for achieving more natural-sounding interactions.
sdyckjq-lab/llm-wiki-skill, with a growth score of 54.84 and 921 stars, offers a personal knowledge base construction skill based on Karpathy's llm-wiki methodology. Its support for multiple platforms has resonated with developers seeking to centralize their knowledge management.
amitshekhariitbhu/llm-internals, sporting a growth score of 43.83 and 588 stars, provides an in-depth exploration of LLM internals, from tokenization to attention to inference optimization. As interest in understanding the inner workings of these complex models grows, this repository is attracting developers eager to learn.
kessler/gemma-gem, with a growth score of 37.22 and 801 stars, enables users to run Google's Gemma 4 model entirely on-device via WebGPU — eliminating the need for API keys, cloud services, or data transmission. Its emphasis on security and self-sufficiency has drawn attention from developers prioritizing data protection.
xoai/sage-wiki, boasting a growth score of 36.97 and 438 stars, compiles papers, articles, and notes into a structured, interlinked wiki using an LLM. Its innovative approach to knowledge organization is resonating with researchers and developers seeking to streamline their information management.
VectifyAI/OpenKB, with a growth score of 32.76 and 445 stars, presents itself as an open LLM knowledge base, offering a centralized hub for storing and retrieving information. As the demand for transparent and accessible knowledge repositories grows, this repository is gaining traction.
lucasastorian/llmwiki, sporting a growth score of 31.29 and 575 stars, provides an open-source implementation of Karpathy's LLM Wiki pattern, allowing users to upload documents and connect their Claude account via MCP. Its flexibility and adaptability have earned it a loyal following among developers.
Pratiyush/llm-wiki, with a growth score of 30.12 and 132 stars, implements the LLM-powered knowledge base concept using Claude Code, Codex CLI, Copilot, Cursor & Gemini sessions. Although smaller in size, this repository is growing steadily as developers explore its potential for centralized information management.
atomicmemory/llm-wiki-compiler, boasting a growth score of 29.72 and 633 stars, compiles raw sources into an interlinked wiki using an LLM, inspired by Karpathy's pattern. Its straightforward approach to knowledge organization has attracted attention from developers seeking efficient solutions.