Today's LLM & Language Models: Fastest-Growing Projects — April 20, 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. Many of these tools are leveraging Karpathy's LLM Wiki pattern to compile and structure knowledge from various sources.
arman-bd/guppylm is a ~9M parameter LLM that talks like a small fish, which has gained significant traction with 2,945 stars and a growth score of 87.05. Its unique approach to generating human-like text has likely contributed to its popularity, as users seek more engaging interactions with language models.
hexiecs/talk-normal is another repository making waves, with 1,284 stars and a growth score of 71.50. This tool makes any LLM talk like a normal person by removing AI-specific jargon and phrasing, which has resonated with developers looking to create more natural-sounding language models.
sdyckjq-lab/llm-wiki-skill is based on Karpathy's llm-wiki methodology for building personal knowledge bases, supporting multiple platforms. With 875 stars and a growth score of 55.87, this repository has seen significant interest from users seeking to compile and structure their knowledge.
amitshekhariitbhu/llm-internals offers a comprehensive guide to learning LLM internals step-by-step, covering topics from tokenization to inference optimization. This educational resource has garnered 554 stars and a growth score of 46.44, as developers seek to deepen their understanding of language models.
xoai/sage-wiki compiles papers, articles, and notes into a structured, interlinked wiki using an LLM. With 427 stars and a growth score of 38.84, this tool has attracted users looking for efficient ways to organize and search their knowledge bases.
kessler/gemma-gem runs Google's Gemma 4 model entirely on-device via WebGPU, eliminating the need for API keys or cloud services. This innovative approach has earned the repository 773 stars and a growth score of 38.67.
Pratiyush/llm-wiki is an implementation of Karpathy's LLM Wiki pattern that allows users to create a knowledge base from Claude Code, Codex CLI, Copilot, Cursor & Gemini sessions. With 125 stars and a growth score of 32.33, this repository has seen interest from developers looking for hands-on experience with the LLM Wiki concept.
lucasastorian/llmwiki is another open-source implementation of Karpathy's LLM Wiki pattern, allowing users to upload documents and connect their Claude account via MCP. This repository has garnered 547 stars and a growth score of 31.53.
askalf/dario is a local LLM router that enables users to access various language model providers through a single endpoint. With 75 stars and a growth score of 31.38, this tool has attracted attention from developers seeking flexibility in their language model integrations.
atomicmemory/llm-wiki-compiler compiles raw sources into an interlinked wiki using the LLM Wiki pattern. This knowledge compiler has earned 614 stars and a growth score of 30.87, as users look for efficient ways to structure and search their knowledge bases.
arman-bd/guppylm is a ~9M parameter LLM that talks like a small fish, which has gained significant traction with 2,945 stars and a growth score of 87.05. Its unique approach to generating human-like text has likely contributed to its popularity, as users seek more engaging interactions with language models.
hexiecs/talk-normal is another repository making waves, with 1,284 stars and a growth score of 71.50. This tool makes any LLM talk like a normal person by removing AI-specific jargon and phrasing, which has resonated with developers looking to create more natural-sounding language models.
sdyckjq-lab/llm-wiki-skill is based on Karpathy's llm-wiki methodology for building personal knowledge bases, supporting multiple platforms. With 875 stars and a growth score of 55.87, this repository has seen significant interest from users seeking to compile and structure their knowledge.
amitshekhariitbhu/llm-internals offers a comprehensive guide to learning LLM internals step-by-step, covering topics from tokenization to inference optimization. This educational resource has garnered 554 stars and a growth score of 46.44, as developers seek to deepen their understanding of language models.
xoai/sage-wiki compiles papers, articles, and notes into a structured, interlinked wiki using an LLM. With 427 stars and a growth score of 38.84, this tool has attracted users looking for efficient ways to organize and search their knowledge bases.
kessler/gemma-gem runs Google's Gemma 4 model entirely on-device via WebGPU, eliminating the need for API keys or cloud services. This innovative approach has earned the repository 773 stars and a growth score of 38.67.
Pratiyush/llm-wiki is an implementation of Karpathy's LLM Wiki pattern that allows users to create a knowledge base from Claude Code, Codex CLI, Copilot, Cursor & Gemini sessions. With 125 stars and a growth score of 32.33, this repository has seen interest from developers looking for hands-on experience with the LLM Wiki concept.
lucasastorian/llmwiki is another open-source implementation of Karpathy's LLM Wiki pattern, allowing users to upload documents and connect their Claude account via MCP. This repository has garnered 547 stars and a growth score of 31.53.
askalf/dario is a local LLM router that enables users to access various language model providers through a single endpoint. With 75 stars and a growth score of 31.38, this tool has attracted attention from developers seeking flexibility in their language model integrations.
atomicmemory/llm-wiki-compiler compiles raw sources into an interlinked wiki using the LLM Wiki pattern. This knowledge compiler has earned 614 stars and a growth score of 30.87, as users look for efficient ways to structure and search their knowledge bases.