Today's AI Frameworks & SDKs: Fastest-Growing Projects — June 26, 2026
Today's the AI Frameworks & SDKs space, we observe a strong focus on modular and scalable solutions that cater to both developers and researchers looking for efficient ways to deploy large language models (LLMs) across distributed systems or integrate advanced features into their projects. CortexPrism's cortex is leading this trend with an impressive growth score, making it clear that there's significant interest in agentic harness systems designed to streamline the development of intelligent applications.
CortexPrism/cortex offers a versatile open-source agentic harness system aimed at developers and researchers working on AI-driven applications. With its high growth score of 48.58 and an increasing number of stars (209), it's evident that CortexPrism is addressing a critical need for more modular and scalable solutions in the AI ecosystem.
leyten/shard is designed to perform pipeline-parallel LLM inference across GPUs on separate machines, making it ideal for large-scale distributed computing environments. Its growth score of 45.18 along with 364 stars suggests that developers are increasingly looking towards shard for efficient and scalable deployment options for their language models.
Simpl3x3/ESEILANE provides a high-performance Knowledge Graph engine tailored for AI, LLMs, and GraphRAG applications, positioning itself as an essential tool for the next generation of intelligent applications. With a growth score of 42.00 and 105 stars, it’s clear that developers are drawn to its promise of enhanced performance in knowledge graph management.
fguzman82/gateGPT converts full transformer models into custom chips, specifically generating names on a Virtex-5 FPGA at an impressive speed of ~56k tokens/second. Its growth score of 39.07 and 572 stars indicate that gateGPT is gaining traction among those interested in hardware-accelerated AI applications and the optimization of transformer models.
Tencent-Hunyuan's UniRL framework for unified multimodal model reinforcement learning has garnered significant attention, with a growth score of 35.08 and 717 stars. This underscores its importance as a comprehensive solution for researchers working on complex reinforcement learning tasks involving multiple modalities.
caezium/Burrow is a free, open-source mole.fit GUI designed to offer users clean, uninstall, optimize, analyze disk, and watch live status functionalities. With a growth score of 34.81 and 787 stars, it’s evident that Burrow is well-regarded for its utility in managing and optimizing macOS environments.
machinefi's trio-retina introduces a model-agnostic state layer for world models, transforming any detector into a standard queryable stream of events and latent states. Its growth score of 30.86 alongside 118 stars highlights the growing demand for versatile tools that can adapt to various detection frameworks.
john-rocky/coreai-model-zoo serves as a community model zoo and knowledge base for Apple Core AI, providing verified on-device models like Qwen3.5 and Gemma 4. With a growth score of 26.72 and 219 stars, it suggests that the repository is valuable for developers seeking to leverage native iOS/macOS capabilities.
OtterMind's Nubase is an open-source, AI-native backend platform designed for AI coding, agentic applications, and modern product teams. Its growth score of 26.58 with 444 stars indicates a strong community interest in its ability to provide memory, database, storage, and authentication services under one umbrella.
Finally, 2aronS/Duel-Agents offers CLI, SDK, and IDE plugins for Duel Agents, enhancing the development experience for those working on agent-based systems. With a growth score of 24.45 and an impressive 1,001 stars, it’s clear that developers are finding value in its comprehensive suite of tools for building and integrating duel agents into their applications.
These projects highlight the diverse needs within the AI community, from hardware optimization to software frameworks, indicating a robust ecosystem driven by innovation and collaboration.
CortexPrism/cortex offers a versatile open-source agentic harness system aimed at developers and researchers working on AI-driven applications. With its high growth score of 48.58 and an increasing number of stars (209), it's evident that CortexPrism is addressing a critical need for more modular and scalable solutions in the AI ecosystem.
leyten/shard is designed to perform pipeline-parallel LLM inference across GPUs on separate machines, making it ideal for large-scale distributed computing environments. Its growth score of 45.18 along with 364 stars suggests that developers are increasingly looking towards shard for efficient and scalable deployment options for their language models.
Simpl3x3/ESEILANE provides a high-performance Knowledge Graph engine tailored for AI, LLMs, and GraphRAG applications, positioning itself as an essential tool for the next generation of intelligent applications. With a growth score of 42.00 and 105 stars, it’s clear that developers are drawn to its promise of enhanced performance in knowledge graph management.
fguzman82/gateGPT converts full transformer models into custom chips, specifically generating names on a Virtex-5 FPGA at an impressive speed of ~56k tokens/second. Its growth score of 39.07 and 572 stars indicate that gateGPT is gaining traction among those interested in hardware-accelerated AI applications and the optimization of transformer models.
Tencent-Hunyuan's UniRL framework for unified multimodal model reinforcement learning has garnered significant attention, with a growth score of 35.08 and 717 stars. This underscores its importance as a comprehensive solution for researchers working on complex reinforcement learning tasks involving multiple modalities.
caezium/Burrow is a free, open-source mole.fit GUI designed to offer users clean, uninstall, optimize, analyze disk, and watch live status functionalities. With a growth score of 34.81 and 787 stars, it’s evident that Burrow is well-regarded for its utility in managing and optimizing macOS environments.
machinefi's trio-retina introduces a model-agnostic state layer for world models, transforming any detector into a standard queryable stream of events and latent states. Its growth score of 30.86 alongside 118 stars highlights the growing demand for versatile tools that can adapt to various detection frameworks.
john-rocky/coreai-model-zoo serves as a community model zoo and knowledge base for Apple Core AI, providing verified on-device models like Qwen3.5 and Gemma 4. With a growth score of 26.72 and 219 stars, it suggests that the repository is valuable for developers seeking to leverage native iOS/macOS capabilities.
OtterMind's Nubase is an open-source, AI-native backend platform designed for AI coding, agentic applications, and modern product teams. Its growth score of 26.58 with 444 stars indicates a strong community interest in its ability to provide memory, database, storage, and authentication services under one umbrella.
Finally, 2aronS/Duel-Agents offers CLI, SDK, and IDE plugins for Duel Agents, enhancing the development experience for those working on agent-based systems. With a growth score of 24.45 and an impressive 1,001 stars, it’s clear that developers are finding value in its comprehensive suite of tools for building and integrating duel agents into their applications.
These projects highlight the diverse needs within the AI community, from hardware optimization to software frameworks, indicating a robust ecosystem driven by innovation and collaboration.