huggingface/transformers
State-of-the-art machine learning library for PyTorch, TensorFlow, and JAX.
~141.0k
Trending GitHub AI repos
and record-breaking papers
A curated, source-linked snapshot of the open-source AI repositories and research reports developers are watching - from DeepSeek and Qwen to frontier agents, reasoning, and evaluation benchmarks.
Trending GitHub AI repositories
A curated snapshot of notable, high-activity open-source AI repositories. Star counts and momentum are reference values that change constantly - open each repo to verify current numbers.
ollama/ollama
Run large language models locally with a simple CLI and REST API.
~124.0k
langchain-ai/langchain
Framework for building LLM applications with composable chains, tools, and agents.
~98.0k
ggml-org/llama.cpp
LLM inference in plain C/C++ with minimal dependencies and broad hardware support.
~72.0k
comfyanonymous/ComfyUI
Powerful and modular diffusion model GUI and backend for image generation.
~63.0k
open-webui/open-webui
User-friendly WebUI for LLMs with a rich feature set and offline support.
~62.0k
langgenius/dify
Open-source LLM app development platform for building AI workflows and agents.
~61.0k
openai/openai-cookbook
Examples and guides for using the OpenAI API, from prompts to function calling.
~61.0k
facebookresearch/segment-anything
Segment Anything Model (SAM): promptable segmentation for images and video.
~49.0k
microsoft/autogen
Multi-agent conversation framework for building next-gen AI applications.
~42.0k
vllm-project/vllm
High-throughput, memory-efficient inference and serving engine for LLMs.
~41.0k
run-llama/llama_index
Data framework for connecting LLMs to your own documents and data sources.
~39.0k
Aider-AI/aider
AI pair programming in your terminal, editing code across your repo.
~26.0k
BerriAI/litellm
Python SDK, proxy server, and gateway for calling 100+ LLMs with one interface.
~16.0k
langchain-ai/langgraph
Low-level orchestration framework for building stateful, multi-agent LLM apps.
~11.0k
In-content ad unit
Top AI research papers & benchmarks
Record-breaking and widely cited research across reasoning, frontier agents, inference efficiency, multimodal systems, and evaluation. Each item links to its official source.
DeepSeek-V3 Technical Report
Why it matters: An open-weights model with a Mixture-of-Experts architecture and a cost-efficient training pipeline that reset expectations for frontier open models.
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
Why it matters: Showed that reasoning can be elicited with reinforcement learning, and open-sourced the reasoning model and its distillation recipes.
Qwen2.5 Technical Report
Why it matters: A family of open models spanning dense and MoE sizes with strong multilingual and coding performance.
Qwen3 Technical Report
Why it matters: Next-generation open model family with hybrid thinking modes and native tool use.
OpenAI o1 System Card
Why it matters: Introduced a model trained with reinforcement learning to think before answering, changing the reasoning paradigm.
Claude 3.7 Sonnet System Card
Why it matters: A hybrid reasoning model with extended thinking plus computer-use tooling for agentic workflows.
Claude 3.5 Sonnet System Card
Why it matters: Introduced computer use, a step toward models that operate tools and interfaces.
Gemini 1.5 Technical Report
Why it matters: A long-context (up to 1M tokens) multimodal model that set a new bar for context handling.
Gemini 2.0
Why it matters: An agentic-era multimodal model with native tool use and real-time capabilities.
The Llama 3 Herd of Models
Why it matters: An open-weight flagship family covering text, vision, and speech, widely used as a baseline.
PagedAttention: Efficient Memory Management for LLM Serving
Why it matters: The memory-management technique behind vLLM that made high-throughput serving practical.
MMLU-Pro
Why it matters: A more challenging, reasoning-heavy successor to MMLU for evaluating knowledge and reasoning.
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
Why it matters: The standard benchmark for measuring how well models can fix real software issues.
HumanEval
Why it matters: A widely cited benchmark for code generation that most coding models report.
LMSYS Chatbot Arena
Why it matters: A crowd-sourced, Elo-based leaderboard for comparing LLMs by human preference.
HELM: Holistic Evaluation of Language Models
Why it matters: A broad, multi-metric evaluation framework for transparent model comparison.
GPQA: A Graduate-Level Google-Proof Q&A Benchmark
Why it matters: A hard, expert-level benchmark used to gauge frontier reasoning.
Gorilla: Large Language Models Connected with Massive APIs
Why it matters: A benchmark and model family for API and tool calling, foundational for agent tool use.
In-content ad unit
Star counts, growth, rankings, and trend status on this page are curated reference snapshots, not live data. They change frequently and should be verified at the source before you rely on them. We do not fetch live numbers without an API key, and we never fabricate figures. Use the "View" and "Source" links on each card to check current values directly on GitHub, arXiv, or the publisher's site.
Trending AI - frequently asked questions
Are the repository star counts live?
No. They are curated reference snapshots labeled with a check date. Star counts change constantly, so verify current numbers on each repository's GitHub page.
Where does the research list come from?
Each paper or report links to its official source - arXiv, the publisher, or the organization's site. The summaries are our concise editorial notes, and items are labeled as official sources or curated commentary.
Can I filter and sort this data?
Yes. Use the search box and category filters to narrow repositories and papers, and sort repositories by stars, name, or momentum.
How often is this refreshed?
The snapshot is curated periodically and each item shows a check date. We intentionally do not claim live data; use the source links for the most current information.