Today's AI Frameworks & SDKs: Fastest-Growing Projects — June 20, 2026
This week, the AI Frameworks & SDKs space continues to see substantial growth and innovation across a variety of projects targeting different aspects of AI development and deployment. CortexPrism's cortex leads the pack with its agentic harness system designed for open-source use, while other notable entries include gateGPT’s custom chip implementation for Transformer models and shard’s pipeline-parallel approach to LLM inference.
CortexPrism/cortex is an open-source agentic harness system aimed at facilitating the development of AI agents. Its recent surge in activity, with a growth score of 64.17 and accumulating over 60 stars, suggests growing interest from developers looking for tools that support agent-centric AI applications.
fguzman82/gateGPT is an ambitious project focused on transforming full Transformer models into custom hardware solutions, specifically targeting Virtex-5 FPGAs with impressive throughput capabilities. The project's high growth score of 62.94 and substantial star count (527) indicate a strong community engagement and interest in edge computing applications for AI.
leyten/shard is designed to enable the efficient distribution of large language model inference tasks across multiple GPUs on separate machines, enhancing performance through pipeline parallelism. With a growth score of 62.40 and 192 stars, shard’s growing popularity can be attributed to its innovative approach in addressing scalability issues for LLMs.
machinefi/trio-retina is an agnostic model layer that converts various detectors into a unified stream of events and latent states, making it highly versatile across different AI applications. The project's growth score of 55.10 and 83 stars reflect its appeal to developers seeking standardized state management for complex systems.
Tencent-Hunyuan/UniRL is a framework dedicated to multimodal model reinforcement learning, aiming to unify various aspects of RL under one roof. With a growth score of 46.17 and 649 stars, UniRL's rapid adoption can be attributed to its comprehensive approach in handling diverse modalities within the realm of reinforcement learning.
caezium/Burrow is a free, open-source GUI for the Mole CLI that provides an intuitive interface for managing macOS system resources and AI agent interactions. The project’s growth score of 43.50 and impressive star count (723) highlight its utility in enhancing user experience through simplified access to complex functionalities.
john-rocky/coreai-model-zoo is a community-driven repository featuring models tailored for Apple's Core AI framework, focusing on iOS/macOS devices. Its growth score of 39.85 and 164 stars suggest that it serves as an invaluable resource for developers working with Apple’s ecosystem and looking to leverage advanced AI capabilities directly on their devices.
study8677/awesome-architecture offers a comprehensive guide to software architecture, including AI-specific components like RAG systems and vector databases. The project's growth score of 39.55 and substantial star count (1,453) underscore its value as an educational resource for both novice and experienced developers aiming to design robust AI-driven applications.
OtterMind/Nubase is an open-source backend platform designed specifically for AI coding, agentic applications, and modern product teams, providing a suite of services including memory management, database support, storage solutions, and authentication. With a growth score of 30.88 and 309 stars, Nubase’s appeal lies in its comprehensive approach to supporting the entire lifecycle of AI-driven application development.
2aronS/Duel-Agents is a suite of tools including CLI, SDK, and IDE plugins for managing duel agents, offering developers an extensive set of utilities for working with agent-based systems. The project's growth score of 27.65 and significant star count (984) indicate its popularity among those interested in developing complex AI-driven applications through a cohesive development environment.
These projects collectively demonstrate the breadth and depth of innovation within the AI frameworks and SDKs space, each addressing unique challenges and offering solutions that are increasingly valuable to developers working on cutting-edge AI applications.
CortexPrism/cortex is an open-source agentic harness system aimed at facilitating the development of AI agents. Its recent surge in activity, with a growth score of 64.17 and accumulating over 60 stars, suggests growing interest from developers looking for tools that support agent-centric AI applications.
fguzman82/gateGPT is an ambitious project focused on transforming full Transformer models into custom hardware solutions, specifically targeting Virtex-5 FPGAs with impressive throughput capabilities. The project's high growth score of 62.94 and substantial star count (527) indicate a strong community engagement and interest in edge computing applications for AI.
leyten/shard is designed to enable the efficient distribution of large language model inference tasks across multiple GPUs on separate machines, enhancing performance through pipeline parallelism. With a growth score of 62.40 and 192 stars, shard’s growing popularity can be attributed to its innovative approach in addressing scalability issues for LLMs.
machinefi/trio-retina is an agnostic model layer that converts various detectors into a unified stream of events and latent states, making it highly versatile across different AI applications. The project's growth score of 55.10 and 83 stars reflect its appeal to developers seeking standardized state management for complex systems.
Tencent-Hunyuan/UniRL is a framework dedicated to multimodal model reinforcement learning, aiming to unify various aspects of RL under one roof. With a growth score of 46.17 and 649 stars, UniRL's rapid adoption can be attributed to its comprehensive approach in handling diverse modalities within the realm of reinforcement learning.
caezium/Burrow is a free, open-source GUI for the Mole CLI that provides an intuitive interface for managing macOS system resources and AI agent interactions. The project’s growth score of 43.50 and impressive star count (723) highlight its utility in enhancing user experience through simplified access to complex functionalities.
john-rocky/coreai-model-zoo is a community-driven repository featuring models tailored for Apple's Core AI framework, focusing on iOS/macOS devices. Its growth score of 39.85 and 164 stars suggest that it serves as an invaluable resource for developers working with Apple’s ecosystem and looking to leverage advanced AI capabilities directly on their devices.
study8677/awesome-architecture offers a comprehensive guide to software architecture, including AI-specific components like RAG systems and vector databases. The project's growth score of 39.55 and substantial star count (1,453) underscore its value as an educational resource for both novice and experienced developers aiming to design robust AI-driven applications.
OtterMind/Nubase is an open-source backend platform designed specifically for AI coding, agentic applications, and modern product teams, providing a suite of services including memory management, database support, storage solutions, and authentication. With a growth score of 30.88 and 309 stars, Nubase’s appeal lies in its comprehensive approach to supporting the entire lifecycle of AI-driven application development.
2aronS/Duel-Agents is a suite of tools including CLI, SDK, and IDE plugins for managing duel agents, offering developers an extensive set of utilities for working with agent-based systems. The project's growth score of 27.65 and significant star count (984) indicate its popularity among those interested in developing complex AI-driven applications through a cohesive development environment.
These projects collectively demonstrate the breadth and depth of innovation within the AI frameworks and SDKs space, each addressing unique challenges and offering solutions that are increasingly valuable to developers working on cutting-edge AI applications.