Today's AI Frameworks & SDKs: Fastest-Growing Projects — May 09, 2026
Today's trends in AI Frameworks & SDKs are dominated by tools that enable efficient deployment and management of large language models (LLMs) on various hardware platforms. The popularity of these tools suggests a growing demand for more accessible and scalable AI solutions. Additionally, we see a surge in projects focused on integrating multiple AI engines and providing unified interfaces.
1bananachicken/MaaNTE takes the top spot with a growth score of 72.04 and over 985 stars. This project provides a powerful assistant driven by the MAAF framework, offering a range of functionalities that have likely contributed to its rapid growth. Its high commit activity in the past 30 days (100 commits) indicates an active development cycle.
HKUDS-AI/polymarket-ai-trading has gained significant attention with a growth score of 66.29 and 139 stars. This repository offers AI-assisted paper trading for Polymarket, utilizing OpenAI market insights, vector search, and Kelly sizing, making it an attractive solution for traders looking to leverage AI in their strategies. Although commit activity is relatively low (4 commits in the past 30 days), its growth suggests a strong interest in AI-driven trading tools.
Warpdot-dev/craft-agents-oss boasts a growth score of 63.62 and 223 stars, making it another notable project this week. This open-source SDK enables the development of multi-engine, multi-LLM applications with a unified interface, which likely appeals to developers seeking more flexibility in their AI projects. Its moderate commit activity (12 commits) indicates ongoing refinement.
Noonghunna/club-3090 has attracted significant attention with a growth score of 62.59 and over 654 stars. This repository provides community-driven recipes for serving LLMs on RTX 3090 hardware, supporting multiple engines like vLLM and llama.cpp. Its high commit activity (100 commits) reflects the active engagement of its community in optimizing AI performance.
Alash3al/stash follows closely with a growth score of 44.73 and 673 stars. This project offers a persistent memory layer for AI agents, storing episodes, facts, and working context in Postgres, which likely appeals to developers seeking more efficient data management solutions. Its high commit activity (97 commits) indicates ongoing development.
Ortera/Cairn has gained notable traction with a growth score of 31.65 and over 780 stars. This state-space search engine, initially validated on autonomous penetration testing, offers a versatile solution for various AI applications. Although commit activity is moderate (50 commits), its high star count suggests significant interest in its capabilities.
TheTom/vllm-swift has secured a growth score of 28.09 and 251 stars with its vLLM Metal plugin powered by mlx-swift, enabling high-performance LLM inference on Apple Silicon hardware. This project's appeal likely stems from its ability to optimize AI performance on specific hardware platforms. Its high commit activity (100 commits) reflects active development.
ZhiYi-R/moon-bridge has a growth score of 24.86 and 69 stars, offering a conversion layer for Anthropic Messages formats to OpenAI Responses API. This tool's growth may be attributed to the increasing demand for seamless integration between different AI engines. Its high commit activity (100 commits) indicates rapid development.
Mininglamp-AI/cider boasts a growth score of 22.42 and 298 stars, providing W8A8/W4A8 inference on Apple Silicon hardware with performance optimizations. This project's appeal likely stems from its ability to unlock faster LLM prefill on specific hardware platforms. Its moderate commit activity (34 commits) reflects ongoing refinement.
Lastly, jegly/Box has a growth score of 21.30 and over 319 stars, offering a private on-device AI suite for Android with support for llama.cpp, whisper.cpp, and other engines. This project's growth may be attributed to its comprehensive set of features and NPU/TPU/GPU acceleration capabilities. Its high commit activity (95 commits) indicates active development.
1bananachicken/MaaNTE takes the top spot with a growth score of 72.04 and over 985 stars. This project provides a powerful assistant driven by the MAAF framework, offering a range of functionalities that have likely contributed to its rapid growth. Its high commit activity in the past 30 days (100 commits) indicates an active development cycle.
HKUDS-AI/polymarket-ai-trading has gained significant attention with a growth score of 66.29 and 139 stars. This repository offers AI-assisted paper trading for Polymarket, utilizing OpenAI market insights, vector search, and Kelly sizing, making it an attractive solution for traders looking to leverage AI in their strategies. Although commit activity is relatively low (4 commits in the past 30 days), its growth suggests a strong interest in AI-driven trading tools.
Warpdot-dev/craft-agents-oss boasts a growth score of 63.62 and 223 stars, making it another notable project this week. This open-source SDK enables the development of multi-engine, multi-LLM applications with a unified interface, which likely appeals to developers seeking more flexibility in their AI projects. Its moderate commit activity (12 commits) indicates ongoing refinement.
Noonghunna/club-3090 has attracted significant attention with a growth score of 62.59 and over 654 stars. This repository provides community-driven recipes for serving LLMs on RTX 3090 hardware, supporting multiple engines like vLLM and llama.cpp. Its high commit activity (100 commits) reflects the active engagement of its community in optimizing AI performance.
Alash3al/stash follows closely with a growth score of 44.73 and 673 stars. This project offers a persistent memory layer for AI agents, storing episodes, facts, and working context in Postgres, which likely appeals to developers seeking more efficient data management solutions. Its high commit activity (97 commits) indicates ongoing development.
Ortera/Cairn has gained notable traction with a growth score of 31.65 and over 780 stars. This state-space search engine, initially validated on autonomous penetration testing, offers a versatile solution for various AI applications. Although commit activity is moderate (50 commits), its high star count suggests significant interest in its capabilities.
TheTom/vllm-swift has secured a growth score of 28.09 and 251 stars with its vLLM Metal plugin powered by mlx-swift, enabling high-performance LLM inference on Apple Silicon hardware. This project's appeal likely stems from its ability to optimize AI performance on specific hardware platforms. Its high commit activity (100 commits) reflects active development.
ZhiYi-R/moon-bridge has a growth score of 24.86 and 69 stars, offering a conversion layer for Anthropic Messages formats to OpenAI Responses API. This tool's growth may be attributed to the increasing demand for seamless integration between different AI engines. Its high commit activity (100 commits) indicates rapid development.
Mininglamp-AI/cider boasts a growth score of 22.42 and 298 stars, providing W8A8/W4A8 inference on Apple Silicon hardware with performance optimizations. This project's appeal likely stems from its ability to unlock faster LLM prefill on specific hardware platforms. Its moderate commit activity (34 commits) reflects ongoing refinement.
Lastly, jegly/Box has a growth score of 21.30 and over 319 stars, offering a private on-device AI suite for Android with support for llama.cpp, whisper.cpp, and other engines. This project's growth may be attributed to its comprehensive set of features and NPU/TPU/GPU acceleration capabilities. Its high commit activity (95 commits) indicates active development.