Today's AI Frameworks & SDKs: Fastest-Growing Projects — September 18, 2026
Today's AI Frameworks & SDKs category highlights a variety of innovative projects ranging from multi-agent systems and distributed computing frameworks to lightweight desktop IDEs for MATLAB scripts, showcasing a diverse set of solutions aimed at advancing the state-of-the-art in artificial intelligence development. The top project this week is AetherLabsAI/RSIAgent, which has seen significant growth with over 300 stars and 37 commits in the last month.
AetherLabsAI/RSIAgent offers a training-free multi-agent framework for recursive self-improvement in new environments through broad-then-deep autonomous exploration and reusable memory. Its impressive growth score of 62.90, along with 305 stars, indicates strong community interest in the concept of self-improving agents that can adapt to novel situations without requiring extensive training data.
0xBakeer/deepseek-v41-flash-spark is a project aimed at optimizing DeepSeek-V4.1-Flash on a single DGX Spark (GB10) system, featuring resident hot experts and NVMe streaming capabilities. With 116 stars and a growth score of 47.00, this repository suggests significant community engagement as developers work to enhance performance through various optimizations and integrations with services like the OpenAI API.
hkqr/my-free-code provides an open-source multi-provider AI gateway for Claude Code and other coding agents, featuring model routing, streaming capabilities, tools support, reasoning fallbacks, and local model support. The project's 31.32 growth score and a notable 637 stars highlight its potential as a versatile tool for developers looking to streamline interactions with multiple AI models across different providers.
Nehanth/swarmllm is an intriguing initiative that splits large language models (LLMs) across devices using WebGPU and WebRTC, enabling peer-to-peer inference. This project's growth score of 28.47 and 326 stars reflect its appeal to developers interested in distributed computing solutions for running massive AI models efficiently.
savellonirourou107-hue/SLX-Studio offers a lightweight desktop IDE specifically designed for MATLAB .m scripts and Simulink .slx models, incorporating visual editing features, plot generation, parameter sweeps, and AI assistance. With 22.66 growth score and 116 stars, this project is gaining traction among users who require an intuitive interface for developing complex simulations and algorithms.
huawei-bayerlab/marigold-v2 revisits diffusion transformers for monocular depth estimation, aiming to improve accuracy in generating depth maps from single images. The repository's growth score of 22.45 and 757 stars indicate a strong interest in advancing the techniques used for this specific application within computer vision.
peonist-ai/halogen-flash-server is focused on optimizing Qwen3.8-Flash-Next performance on Strix Halo (gfx1151) hardware, aiming to deliver high-speed execution and efficient resource utilization. With 21.96 growth score and 527 stars, the project demonstrates a significant community interest in enhancing AI model performance through specialized hardware optimizations.
TokenRhythm/NeoHorse introduces NeoHorse-1, an approach towards recursive self-improvement via agentic post-training with routing harnesses. The project's growth score of 21.21 and 516 stars reflect a growing community interest in developing intelligent agents capable of continuous learning and adaptation.
lucidrains/RLT presents an implementation of the recurrent looped transformer, a novel architecture proposed by Yifan Zhang from Princeton University. With a growth score of 20.80 and 52 stars, this project is attracting attention from researchers and developers interested in exploring new architectures for sequence modeling tasks.
Finally, TrenTorch/TrenTorch offers an educational framework inspired by Harvard's TinyTorch that allows learners to build their own PyTorch-like library from scratch. The growth score of 17.84 and 303 stars indicate the project's potential as a valuable learning resource for those interested in understanding the inner workings of deep learning frameworks.
Today's selection of projects highlights the breadth and depth of innovation within the AI Frameworks & SDKs space, with each tool addressing unique challenges and opportunities in areas such as multi-agent systems, distributed computing, model optimization, and educational tools.
AetherLabsAI/RSIAgent offers a training-free multi-agent framework for recursive self-improvement in new environments through broad-then-deep autonomous exploration and reusable memory. Its impressive growth score of 62.90, along with 305 stars, indicates strong community interest in the concept of self-improving agents that can adapt to novel situations without requiring extensive training data.
0xBakeer/deepseek-v41-flash-spark is a project aimed at optimizing DeepSeek-V4.1-Flash on a single DGX Spark (GB10) system, featuring resident hot experts and NVMe streaming capabilities. With 116 stars and a growth score of 47.00, this repository suggests significant community engagement as developers work to enhance performance through various optimizations and integrations with services like the OpenAI API.
hkqr/my-free-code provides an open-source multi-provider AI gateway for Claude Code and other coding agents, featuring model routing, streaming capabilities, tools support, reasoning fallbacks, and local model support. The project's 31.32 growth score and a notable 637 stars highlight its potential as a versatile tool for developers looking to streamline interactions with multiple AI models across different providers.
Nehanth/swarmllm is an intriguing initiative that splits large language models (LLMs) across devices using WebGPU and WebRTC, enabling peer-to-peer inference. This project's growth score of 28.47 and 326 stars reflect its appeal to developers interested in distributed computing solutions for running massive AI models efficiently.
savellonirourou107-hue/SLX-Studio offers a lightweight desktop IDE specifically designed for MATLAB .m scripts and Simulink .slx models, incorporating visual editing features, plot generation, parameter sweeps, and AI assistance. With 22.66 growth score and 116 stars, this project is gaining traction among users who require an intuitive interface for developing complex simulations and algorithms.
huawei-bayerlab/marigold-v2 revisits diffusion transformers for monocular depth estimation, aiming to improve accuracy in generating depth maps from single images. The repository's growth score of 22.45 and 757 stars indicate a strong interest in advancing the techniques used for this specific application within computer vision.
peonist-ai/halogen-flash-server is focused on optimizing Qwen3.8-Flash-Next performance on Strix Halo (gfx1151) hardware, aiming to deliver high-speed execution and efficient resource utilization. With 21.96 growth score and 527 stars, the project demonstrates a significant community interest in enhancing AI model performance through specialized hardware optimizations.
TokenRhythm/NeoHorse introduces NeoHorse-1, an approach towards recursive self-improvement via agentic post-training with routing harnesses. The project's growth score of 21.21 and 516 stars reflect a growing community interest in developing intelligent agents capable of continuous learning and adaptation.
lucidrains/RLT presents an implementation of the recurrent looped transformer, a novel architecture proposed by Yifan Zhang from Princeton University. With a growth score of 20.80 and 52 stars, this project is attracting attention from researchers and developers interested in exploring new architectures for sequence modeling tasks.
Finally, TrenTorch/TrenTorch offers an educational framework inspired by Harvard's TinyTorch that allows learners to build their own PyTorch-like library from scratch. The growth score of 17.84 and 303 stars indicate the project's potential as a valuable learning resource for those interested in understanding the inner workings of deep learning frameworks.
Today's selection of projects highlights the breadth and depth of innovation within the AI Frameworks & SDKs space, with each tool addressing unique challenges and opportunities in areas such as multi-agent systems, distributed computing, model optimization, and educational tools.