Trending model & inference repos on GitHub today

As of Oct 22, 2026, the GitHub model & inference repos that gained the most stars on Oct 21, 2026 is deepseek-ai/DeepInfer (+3,838).

Watchlist	#	Repository	Stars gained	Total stars	Site traffic
☆	1in	deepseek-ai/DeepInferOpen-source offline inference framework for DeepSeek models with optimized kernels, parallelism, and scheduling.+3,838 stars in the last day · Python · created 2026-10	+3,838	18,843	—
☆	2in	Comfy-Org/ComfyUINode-based GUI and backend that runs diffusion models locally for image, video, audio and 3D generation.+769 stars in the last day · Python · created 2023-01	+769	141,762	3.9M · 2026-09
☆	3in	Niko1221/StrataStrata is a local inference engine running Qwen3.8-Flash-Next on consumer hardware.+744 stars in the last day · C++ · created 2026-09	+744	30,133	—
☆	4in	tashfeenahmed/freellmapiAggregates free tiers from 34 LLM providers into one OpenAI-compatible endpoint with routing and failover.+674 stars in the last day · TypeScript · created 2026-04	+674	39,146	traffic →
☆	5in	Fractal-Compute-Orchestrations/FractalWorkspaceA distributed framework for federated learning and edge model training across Android and Python nodes.+671 stars in the last day · Python · created 2026-05	+671	3,101	traffic →
☆	6in	Fractal-Compute-Orchestrations/FractalAndroidAndroid edge compute client running on-device federated learning and distributed ML workloads.+666 stars in the last day · Kotlin · created 2025-11	+666	3,010	traffic →
☆	7in	headroomlabs-ai/headroomCompresses tool outputs, logs, files, and RAG chunks locally before they reach the LLM, reducing token usage.+564 stars in the last day · Python · created 2026-01	+564	77,434	—
☆	8in	vllm-project/vllmvLLM is a high-throughput, memory-efficient open-source engine for LLM inference and serving.+501 stars in the last day · Python · created 2023-02	+501	95,038	traffic →
☆	9in	rohitg00/ai-engineering-from-scratchOpen-source AI engineering curriculum with 542 lessons covering model internals, retrieval, and agent runtimes.+329 stars in the last day · Python · created 2026-03	+329	72,204	traffic →
☆	10in	Alishahryar1/free-claude-codeA proxy that routes Claude Code and other coding agents to 61 free and paid model providers.+259 stars in the last day · Python · created 2026-01	+259	62,510	—
☆	11in	lyogavin/airllmAirLLM enables inference of very large LLMs on a single low-VRAM GPU via layer-wise loading.+240 stars in the last day · Jupyter Notebook · created 2023-06	+240	40,212	—
☆	12in	ollama/ollamaOllama is a tool for running and managing open LLMs locally or in the cloud.+139 stars in the last day · Go · created 2023-06	+139	184,246	10.2M · 2026-08
☆	13in	p-e-w/hereticHeretic automatically removes refusal behavior from transformer language models using abliteration and Optuna-based parameter search.+125 stars in the last day · Python · created 2025-09	+125	36,155	—
☆	14in	Tencent/WeKnoraOpen-source LLM knowledge platform combining RAG retrieval, a reasoning agent, and a self-maintaining wiki over documents.+123 stars in the last day · Go · created 2025-07	+123	35,030	—
☆	15in	docling-project/doclingDocling parses PDFs and other document formats into structured data for generative AI and RAG pipelines.+103 stars in the last day · Python · created 2024-07	+103	69,333	—
☆	16in	mnfst/awesome-free-llm-apisCurated list of LLM providers offering permanent free API tiers and keys.+99 stars in the last day · JavaScript · created 2026-03	+99	10,934	—
☆	17in	router-for-me/CLIProxyAPIProxy server exposing CLI accounts as OpenAI, Gemini, and Claude compatible API endpoints.+98 stars in the last day · Go · created 2025-07	+98	55,941	—
☆	18in	microsoft/generative-ai-for-beginnersMicrosoft's 21-lesson open curriculum teaching fundamentals of building generative AI applications.+92 stars in the last day · Jupyter Notebook · created 2023-06	+92	121,868	—
☆	19in	Wei-Shaw/sub2apiGo/Vue AI API gateway relaying Claude, OpenAI, Gemini and Grok subscriptions into one endpoint.+91 stars in the last day · Go · created 2025-12	+91	44,574	—
☆	20in	QuantumNous/new-apiSelf-hosted AI gateway aggregating multiple LLM providers into OpenAI, Claude, and Gemini-compatible APIs.+88 stars in the last day · Go · created 2023-11	+88	50,456	traffic →
☆	21in	BerriAI/litellmOpen-source AI gateway and Python SDK that unifies calls to 100+ LLM providers with cost tracking and load balancing.+86 stars in the last day · Python · created 2023-07	+86	61,779	—
☆	22in	rasbt/LLMs-from-scratchCode for building, pretraining, and finetuning a GPT-like LLM in PyTorch from scratch.+82 stars in the last day · Jupyter Notebook · created 2023-07	+82	107,079	traffic →
☆	23in	chatanywhere/GPT_API_freeProvides free and paid API access to GPT, DeepSeek, Claude, Gemini and other LLMs.+70 stars in the last day · created 2023-04	+70	44,647	—
☆	24in	infiniflow/ragflowRAGFlow is an open-source RAG engine combining retrieval-augmented generation with agent capabilities for LLM context.+64 stars in the last day · Go · created 2023-12	+64	92,635	traffic →
☆	25in	decolua/9routerA local router that connects AI coding tools to 40+ LLM providers with fallback and token compression.+63 stars in the last day · JavaScript · created 2026-01	+63	31,197	traffic →
☆	26in	datawhalechina/all-in-ragA Chinese RAG tutorial covering data preparation, embeddings, vector databases, retrieval, and evaluation.+62 stars in the last day · Python · created 2025-06	+62	12,437	—
☆	27in	multimodal-art-projection/YuEYuE2 generates songs from lyrics and style prompts using symbolic planning and audio rendering.+61 stars in the last day · Python · created 2025-01	+61	12,508	—
☆	28in	unclecode/crawl4aiOpen-source web crawler that converts websites into clean, LLM-ready Markdown for RAG and AI agents.+57 stars in the last day · Python · created 2024-05	+57	86,007	traffic →
☆	29in	sgl-project/sglangSGLang is an open-source serving framework for efficient inference of large language, multimodal, and diffusion models.+55 stars in the last day · Python · created 2024-01	+55	37,593	traffic →
☆	30in	mlabonne/llm-courseAn educational course with roadmaps and Colab notebooks covering LLM fundamentals, training, and application deployment.+53 stars in the last day · created 2023-06	+53	84,580	—
☆	31in	langfuse/langfuseLangfuse is an open-source LLM engineering platform for tracing, evaluating, and managing prompts in AI applications.+49 stars in the last day · TypeScript · created 2023-05	+49	36,165	traffic →
☆	32in	opendataloader-project/opendataloader-pdfOpen-source PDF parser extracting Markdown, JSON, and HTML for AI and RAG pipelines, plus PDF accessibility tagging.+49 stars in the last day · Java · created 2025-05	+49	29,800	traffic →
☆	33in	lidge-jun/opencodexLocal proxy that lets Codex, Claude Code and similar clients run any LLM provider.+49 stars in the last day · TypeScript · created 2026-06	+49	17,756	—
☆	34in	MakazhanAlpamys/SoupCLI tool that fine-tunes LLMs from a single YAML config with layer streaming.+48 stars in the last day · Python · created 2026-02	+48	9,540	traffic →
☆	35in	llm-d/llm-dllm-d is a Kubernetes distributed inference serving stack for optimizing LLM deployment performance.+45 stars in the last day · Python · created 2025-04	+45	4,940	traffic →
☆	36in	VectifyAI/PageIndexVectorless RAG engine that builds tree indexes of documents and retrieves via LLM reasoning.+43 stars in the last day · Python · created 2025-04	+43	39,445	traffic →
☆	37in	topoteretes/cogneeCognee is an open-source memory layer that builds knowledge graphs from documents for AI agents.+42 stars in the last day · Python · created 2023-08	+42	32,648	traffic →
☆	38in	PaddlePaddle/PaddleOCRPaddleOCR is a lightweight OCR and document parsing toolkit that converts PDFs and images into structured data for LLMs.+41 stars in the last day · Python · created 2020-05	+41	91,497	traffic →
☆	39in	hiyouga/LlamaFactoryUnified framework for efficient fine-tuning of over 100 LLMs and VLMs.+39 stars in the last day · Python · created 2023-05	+39	75,650	—
☆	40in	promptfoo/promptfooPromptfoo is an open-source CLI and library for evaluating, comparing, and red-teaming LLM prompts and applications.+39 stars in the last day · TypeScript · created 2023-04	+39	26,257	traffic →
☆	41in	bojieli/ai-infra-bookOpen-source book quantitatively deriving LLM inference and training system design from hardware constraints and model architectures.+39 stars in the last day · Python · created 2026-08	+39	6,850	—
☆	42in	cactus-compute/needleNeedle is a tiny 2-bit on-device foundation model for tool calls, extraction, embeddings and speech.+38 stars in the last day · Python · created 2026-02	+38	14,191	—
☆	43in	AlexsJones/llmfitCLI tool that detects hardware and recommends local LLMs that fit your RAM, CPU, and GPU.+36 stars in the last day · Rust · created 2026-02	+36	38,213	—
☆	44in	mudler/LocalAILocalAI is an open-source AI engine running LLMs, vision, voice, image and video models on any hardware.+35 stars in the last day · Go · created 2023-03	+35	49,847	traffic →
☆	45in	huggingface/transformersHugging Face's model-definition framework for training and inference of text, vision, audio, and multimodal models.+34 stars in the last day · Python · created 2018-10	+34	167,965	—
☆	46in	SYSTRAN/faster-whisperA CTranslate2-based reimplementation of OpenAI Whisper enabling faster, lower-memory speech transcription with quantization.+33 stars in the last day · Python · created 2023-02	+33	26,038	—
☆	47in	comet-ml/opikOpen-source LLM observability and evaluation platform for tracing agents, running evaluations, and monitoring production.+33 stars in the last day · Python · created 2023-05	+33	22,822	traffic →
☆	48in	maximhq/bifrostOpen-source AI gateway unifying 23+ LLM providers behind one OpenAI-compatible API with failover, caching and MCP support.+32 stars in the last day · Go · created 2025-03	+32	9,011	traffic →
☆	49in	fishaudio/fish-speechFish Speech is an open-source TTS system with pretrained models, inference and API server.+27 stars in the last day · Python · created 2023-10	+27	33,352	—
☆	50in	harvard-edge/cs249r_bookHarvard CS249r open textbook series on machine learning systems, with labs, TinyTorch, and hardware kits.+26 stars in the last day · Python · created 2023-09	+26	29,112	traffic →

Data: daily star-total snapshots from the GitHub GraphQL API; stars gained = difference between snapshots. GitHub search and GH Archive are used to find repos. Updated daily around 02:30 UTC, counting through the last complete UTC day. · By Kai

How we rank · Weekly archive · RSS · JSON API