Today's AI Frameworks & SDKs: Fastest-Growing Projects — May 16, 2026
Today's AI Frameworks & SDKs, we see continued momentum with several projects gaining traction as developers and researchers explore new ways to optimize large language models (LLMs) for inference and deployment. Among the standout tools is lightseekorg/tokenspeed, which has seen significant growth, likely due to its innovative approach to speeding up LLM inference.
lightseekorg/tokenspeed: TokenSpeed is a speed-of-light LLM inference engine designed to optimize performance during model execution. With a high growth score of 95 and over a thousand stars on GitHub, it's clear that developers are interested in the potential for faster and more efficient AI applications.
1bananachicken/MaaNTE: This project is an automation assistant for Neverness to Everness, featuring auto-fishing, auto-selling fish, and other automated tasks. Its strong community engagement and high star count (1,287) indicate a robust user base interested in enhancing their gaming experience through AI.
noonghunna/club-3090: Club 3090 provides community recipes for serving large language models on RTX 3090 GPUs, supporting multiple engines and model configurations. With nearly a thousand stars and a steady stream of commits (100 in the last month), this repository is gaining recognition among developers looking to optimize LLM performance.
TheRunicDev/MaaNTE: This project offers automation features for Neverness to Everness, including auto-fishing and daily reward claiming. Despite its lower growth score compared to others, it has garnered 296 stars, suggesting a dedicated user community interested in enhancing gameplay efficiency through AI tools.
oritera/Cairn: Cairn is a general-purpose state-space search engine validated on autonomous penetration testing tasks. Its moderate growth and star count (31.09 and 995 respectively) indicate steady interest from security researchers looking to leverage advanced search algorithms for automated testing.
alash3al/stash: Stash provides a persistent memory layer for AI agents, enabling them to store episodes, facts, and working context in Postgres. With 691 stars and frequent commits (97 in the last month), this project is attracting developers interested in creating more sophisticated and self-aware AI systems.
iFurySt/open-codex-computer-use: This tool offers an open-source alternative to Codex Computer Use, providing a platform for users to manage their digital environments through automation. Its high star count (789) reflects the demand for accessible tools that enhance productivity and streamline computer usage.
Ontos-AI/knowhere: Knowhere is designed to extract, parse, and output structured data ready for AI agents and retrieval-augmented generation (RAG). With 100 commits in the past month and a growing community of 187 stars, this project demonstrates strong interest from developers looking to integrate structured data into AI workflows.
wanghuan9/skill-manager: SkillDock is a desktop control center for managing AI skills and MCP servers with Git-aware updates. Despite its smaller star count (71), the tool's focus on integration with popular development tools like GitHub Copilot suggests it addresses specific needs within niche communities of developers.
DioCrafts/OpenFoundry: OpenFoundry aims to be an open-source alternative to Palantir Foundry, offering a comprehensive platform for data management and AI-driven decision-making. With 100 commits in the last month and growing interest (190 stars), this project is attracting attention from developers seeking self-hosted solutions for enterprise-level data operations.
Today's trends highlight a diverse array of projects catering to different needs within the broader AI ecosystem, from optimizing LLM performance to enhancing automation and data management capabilities.
lightseekorg/tokenspeed: TokenSpeed is a speed-of-light LLM inference engine designed to optimize performance during model execution. With a high growth score of 95 and over a thousand stars on GitHub, it's clear that developers are interested in the potential for faster and more efficient AI applications.
1bananachicken/MaaNTE: This project is an automation assistant for Neverness to Everness, featuring auto-fishing, auto-selling fish, and other automated tasks. Its strong community engagement and high star count (1,287) indicate a robust user base interested in enhancing their gaming experience through AI.
noonghunna/club-3090: Club 3090 provides community recipes for serving large language models on RTX 3090 GPUs, supporting multiple engines and model configurations. With nearly a thousand stars and a steady stream of commits (100 in the last month), this repository is gaining recognition among developers looking to optimize LLM performance.
TheRunicDev/MaaNTE: This project offers automation features for Neverness to Everness, including auto-fishing and daily reward claiming. Despite its lower growth score compared to others, it has garnered 296 stars, suggesting a dedicated user community interested in enhancing gameplay efficiency through AI tools.
oritera/Cairn: Cairn is a general-purpose state-space search engine validated on autonomous penetration testing tasks. Its moderate growth and star count (31.09 and 995 respectively) indicate steady interest from security researchers looking to leverage advanced search algorithms for automated testing.
alash3al/stash: Stash provides a persistent memory layer for AI agents, enabling them to store episodes, facts, and working context in Postgres. With 691 stars and frequent commits (97 in the last month), this project is attracting developers interested in creating more sophisticated and self-aware AI systems.
iFurySt/open-codex-computer-use: This tool offers an open-source alternative to Codex Computer Use, providing a platform for users to manage their digital environments through automation. Its high star count (789) reflects the demand for accessible tools that enhance productivity and streamline computer usage.
Ontos-AI/knowhere: Knowhere is designed to extract, parse, and output structured data ready for AI agents and retrieval-augmented generation (RAG). With 100 commits in the past month and a growing community of 187 stars, this project demonstrates strong interest from developers looking to integrate structured data into AI workflows.
wanghuan9/skill-manager: SkillDock is a desktop control center for managing AI skills and MCP servers with Git-aware updates. Despite its smaller star count (71), the tool's focus on integration with popular development tools like GitHub Copilot suggests it addresses specific needs within niche communities of developers.
DioCrafts/OpenFoundry: OpenFoundry aims to be an open-source alternative to Palantir Foundry, offering a comprehensive platform for data management and AI-driven decision-making. With 100 commits in the last month and growing interest (190 stars), this project is attracting attention from developers seeking self-hosted solutions for enterprise-level data operations.
Today's trends highlight a diverse array of projects catering to different needs within the broader AI ecosystem, from optimizing LLM performance to enhancing automation and data management capabilities.