Today's AI Frameworks & SDKs: Fastest-Growing Projects — May 09, 2026
Today's the AI Frameworks & SDKs space, we're seeing a surge in tools focused on serving large language models (LLMs) efficiently and privately. Repositories that enable LLM inference on specialized hardware, such as Apple Silicon and RTX 3090, are gaining traction. Additionally, projects providing frameworks for building AI-powered applications with multiple engines and model-agnostic support are also experiencing significant growth.
MaaNTE, a MAAF framework-driven tool with a Growth Score of 73.88 and over 1,015 stars, is growing rapidly due to its ability to provide a robust platform for building AI applications. Its high commit activity in the past 30 days (100 commits) indicates active development and community engagement.
The polymarket-ai-trading repository by HKUDS-AI has a Growth Score of 66.29 and 139 stars, with only 4 commits in the past month. Despite the relatively low commit activity, its AI-assisted paper trading capabilities for Polymarket are likely driving interest among developers looking to leverage OpenAI market insights.
Noonghunna's club-3090 has a Growth Score of 63.77 and an impressive 674 stars, with 100 commits in the past month. This community-driven project provides recipes for serving LLMs on RTX 3090, making it an attractive resource for developers seeking to optimize their AI workflows.
Warpdot-dev's craft-agents-oss boasts a Growth Score of 63.44 and 220 stars, with 12 commits in the past month. Its multi-engine support (including vLLM and llama.cpp) and model-agnostic design make it an appealing choice for developers building complex AI applications.
Alash3al's stash has a Growth Score of 44.73 and 673 stars, with an impressive 97 commits in the past month. This persistent memory layer for AI agents provides episodes, facts, and working context storage in Postgres, making it a valuable resource for developers seeking to improve their AI model performance.
Oritera's Cairn has a Growth Score of 32.50 and 796 stars, with 50 commits in the past month. As a general-purpose state-space search engine validated on autonomous penetration testing, Cairn is likely attracting interest from developers working on complex problem-solving applications.
TheTom's vllm-swift has a Growth Score of 28.09 and 251 stars, with an impressive 100 commits in the past month. This high-performance LLM inference plugin for Apple Silicon (powered by mlx-swift) is growing rapidly due to its ability to unlock efficient AI processing on Apple devices.
ZhiYi-R's moon-bridge has a Growth Score of 25.32 and 79 stars, with an impressive 100 commits in the past month. This conversion layer for Anthropic Messages formats to Codex-compatible OpenAI Responses API is likely gaining traction among developers seeking to integrate multiple AI engines into their applications.
Mininglamp-AI's cider has a Growth Score of 23.54 and 313 stars, with 36 commits in the past month. Its optimized W8A8/W4A8 inference on Apple Silicon unlocks faster LLM prefill performance, making it an attractive resource for developers seeking to improve AI efficiency on Apple devices.
Jegly's Box has a Growth Score of 21.84 and 319 stars, with 99 commits in the past month. This private on-device AI suite for Android provides a range of features, including llama.cpp, whisper.cpp, and NPU/TPU/GPU acceleration, making it a valuable resource for developers seeking to build secure and efficient AI-powered applications.
MaaNTE, a MAAF framework-driven tool with a Growth Score of 73.88 and over 1,015 stars, is growing rapidly due to its ability to provide a robust platform for building AI applications. Its high commit activity in the past 30 days (100 commits) indicates active development and community engagement.
The polymarket-ai-trading repository by HKUDS-AI has a Growth Score of 66.29 and 139 stars, with only 4 commits in the past month. Despite the relatively low commit activity, its AI-assisted paper trading capabilities for Polymarket are likely driving interest among developers looking to leverage OpenAI market insights.
Noonghunna's club-3090 has a Growth Score of 63.77 and an impressive 674 stars, with 100 commits in the past month. This community-driven project provides recipes for serving LLMs on RTX 3090, making it an attractive resource for developers seeking to optimize their AI workflows.
Warpdot-dev's craft-agents-oss boasts a Growth Score of 63.44 and 220 stars, with 12 commits in the past month. Its multi-engine support (including vLLM and llama.cpp) and model-agnostic design make it an appealing choice for developers building complex AI applications.
Alash3al's stash has a Growth Score of 44.73 and 673 stars, with an impressive 97 commits in the past month. This persistent memory layer for AI agents provides episodes, facts, and working context storage in Postgres, making it a valuable resource for developers seeking to improve their AI model performance.
Oritera's Cairn has a Growth Score of 32.50 and 796 stars, with 50 commits in the past month. As a general-purpose state-space search engine validated on autonomous penetration testing, Cairn is likely attracting interest from developers working on complex problem-solving applications.
TheTom's vllm-swift has a Growth Score of 28.09 and 251 stars, with an impressive 100 commits in the past month. This high-performance LLM inference plugin for Apple Silicon (powered by mlx-swift) is growing rapidly due to its ability to unlock efficient AI processing on Apple devices.
ZhiYi-R's moon-bridge has a Growth Score of 25.32 and 79 stars, with an impressive 100 commits in the past month. This conversion layer for Anthropic Messages formats to Codex-compatible OpenAI Responses API is likely gaining traction among developers seeking to integrate multiple AI engines into their applications.
Mininglamp-AI's cider has a Growth Score of 23.54 and 313 stars, with 36 commits in the past month. Its optimized W8A8/W4A8 inference on Apple Silicon unlocks faster LLM prefill performance, making it an attractive resource for developers seeking to improve AI efficiency on Apple devices.
Jegly's Box has a Growth Score of 21.84 and 319 stars, with 99 commits in the past month. This private on-device AI suite for Android provides a range of features, including llama.cpp, whisper.cpp, and NPU/TPU/GPU acceleration, making it a valuable resource for developers seeking to build secure and efficient AI-powered applications.