Open Source

Tencent's open-source 3D asset generation system with 13k+ GitHub stars, creating high-resolution textured 3D models from a single image using a two-stage diffusion pipeline.

Open Source113

TrendingLLM

GitHub

4.2K431

GLM-4.5

Z.AI (Zhipu AI)

Z.AI's open-source MoE LLM series with 355B parameters, achieving top-3 benchmark rankings with agentic coding, hybrid reasoning modes, and 128K-200K context support under Apache-2.0.

TimesFM

Google Research

Google Research's pretrained time-series foundation model with decoder-only architecture, achieving zero-shot forecasting across domains with 200M parameters and 16k context length.

Graphiti

Zep

Open-source framework for building real-time, temporally-aware knowledge graphs for AI agents with hybrid search, bi-temporal tracking, and sub-second query latency.

RAG-Anything

HKUDS (Hong Kong University of Science and Technology)

All-in-one multimodal RAG framework that seamlessly processes text, images, tables, equations, and charts through a unified knowledge graph pipeline.

Agent Lightning

Microsoft

Microsoft's open-source framework for training and optimizing any AI agent using reinforcement learning, prompt optimization, and fine-tuning with minimal code changes.

Exo

exo labs

Open-source distributed AI cluster framework that connects multiple devices to run frontier LLMs locally, with automatic device discovery and Thunderbolt 5 RDMA support.

nanochat

Andrej Karpathy

Andrej Karpathy's minimal LLM training framework that enables end-to-end ChatGPT-class model training from scratch for approximately $100.

Unsloth

Unsloth AI

Open-source LLM fine-tuning optimizer delivering 2x faster training with 70% less VRAM, supporting models from GPT-OSS to DeepSeek with zero accuracy loss.

Machine Learning Systems (CS249r)

harvard-edge

Open-source Harvard ML systems textbook with TinyTorch framework, edge deployment labs, and multi-language support, trending with 19.7K GitHub stars.

Kotaemon

Cinnamon

Open-source RAG framework for document QA with hybrid retrieval, multi-modal parsing, and multi-user support, backed by 25K+ GitHub stars.

InferenceX

SemiAnalysisAI

Open-source continuous inference benchmarking platform by SemiAnalysis that tests AI frameworks nightly across NVIDIA and AMD frontier GPUs.

Apache-2.097

17 18 19 20 21 22 23 24 25

310 projects

Sort:

Trending3D

GitHub

13.2K1.3K

Hunyuan3D-2

Tencent Hunyuan

Tencent's open-source 3D asset generation system with 13k+ GitHub stars, creating high-resolution textured 3D models from a single image using a two-stage diffusion pipeline.

GLM-4.5

Z.AI (Zhipu AI)

Z.AI's open-source MoE LLM series with 355B parameters, achieving top-3 benchmark rankings with agentic coding, hybrid reasoning modes, and 128K-200K context support under Apache-2.0.

TimesFM

Google Research

Google Research's pretrained time-series foundation model with decoder-only architecture, achieving zero-shot forecasting across domains with 200M parameters and 16k context length.

Graphiti

Zep

Open-source framework for building real-time, temporally-aware knowledge graphs for AI agents with hybrid search, bi-temporal tracking, and sub-second query latency.

RAG-Anything

HKUDS (Hong Kong University of Science and Technology)

All-in-one multimodal RAG framework that seamlessly processes text, images, tables, equations, and charts through a unified knowledge graph pipeline.

Agent Lightning

Microsoft

Microsoft's open-source framework for training and optimizing any AI agent using reinforcement learning, prompt optimization, and fine-tuning with minimal code changes.

Exo

exo labs

Open-source distributed AI cluster framework that connects multiple devices to run frontier LLMs locally, with automatic device discovery and Thunderbolt 5 RDMA support.

nanochat

Andrej Karpathy

Andrej Karpathy's minimal LLM training framework that enables end-to-end ChatGPT-class model training from scratch for approximately $100.

Unsloth

Unsloth AI

Open-source LLM fine-tuning optimizer delivering 2x faster training with 70% less VRAM, supporting models from GPT-OSS to DeepSeek with zero accuracy loss.