People to follow on X
A hand-picked list of X accounts worth following for local models, local RAG and document Q&A, local inference, and NVIDIA DGX Spark. Every account was checked as active (posting in the last 60 days) and on topic.
Model builders and labs · Local inference tooling · RAG and document tooling · Researchers and benchmarkers · DGX Spark and home-lab community
Model builders and labs
| Account | Why follow | Followers |
|---|---|---|
| @Alibaba_Qwen Qwen | Official Qwen account: announcements for Qwen open-weight model releases. | 298.7K |
| @deepseek_ai DeepSeek | Official DeepSeek account: DeepSeek model and paper releases. | 1.2M |
| @Zai_org Z.ai | The lab behind GLM, including GLM-5.3-Flash (our #1 local model). | 167.4K |
| @Kimi_Moonshot Kimi.ai | Moonshot AI's account for Kimi model releases. | 382.7K |
| @MiniMax_AI MiniMax (official) | MiniMax releases; recently posting about M3.1 Flash and its vLLM/SGLang launch partners. | 128.7K |
| @MistralAI Mistral AI | Mistral's open and commercial model releases. | 225.1K |
| @NousResearch Nous Research | Open models and the Hermes Agent; posts frequent Hermes plugin and integration news. | 287.9K |
| @huggingface Hugging Face | The hub where most open-weight models and GGUFs get published. | 827.6K |
Local inference tooling
| Account | Why follow | Followers |
|---|---|---|
| @ggerganov Georgi Gerganov | Creator of llama.cpp and ggml, the engine under many local apps. | 73.4K |
| @ollama ollama | Ollama release news and new models you can pull and run locally. | 186.7K |
| @vllm_project vLLM | vLLM serving engine; posts Day-0 support for new open models with serve commands. | 53.0K |
| @sgl_project SGLang | SGLang serving engine; release notes and Day-0 model support. | 12.0K |
| @lmstudio LM Studio | LM Studio desktop app updates for running models locally. | 66.8K |
| @UnslothAI Unsloth AI | Unsloth Dynamic GGUFs and fine-tuning; posts how much RAM new models need locally. | 102.1K |
| @exolabs EXO Labs | EXO: open-source software for running large models across your own devices. | 54.8K |
| @awnihannun Awni Hannun | Works on Apple's MLX framework for running models on Macs. | 45.1K |
RAG and document tooling
| Account | Why follow | Followers |
|---|---|---|
| @jerryjliu0 Jerry Liu | LlamaIndex co-founder/CEO; posts about document parsing (LlamaParse, OCRia-1). | 86.7K |
| @llama_index LlamaIndex 🦙 | LlamaIndex: document parsing and RAG framework news. | 121.8K |
| @AnythingLLM AnythingLLM | All-in-one desktop app for private RAG with any LLM. | 3,288 |
| @tcarambat Tim Carambat | Builder of AnythingLLM. | 3,461 |
| @OpenWebUI Open WebUI | Open WebUI, the self-hosted chat UI with built-in document RAG. | 11.8K |
| @infiniflowai InfiniFlow | InfiniFlow, makers of the RAGFlow open-source RAG engine. | 1,972 |
Researchers and benchmarkers
| Account | Why follow | Followers |
|---|---|---|
| @ArtificialAnlys Artificial Analysis | Independent model benchmarks and leaderboards, posted near-daily. | 164.6K |
| @simonw Simon Willison | Simon Willison: hands-on notes on new models and running LLMs locally (creator of the llm CLI). | 244.3K |
| @rasbt Sebastian Raschka | Sebastian Raschka: LLM research explainers; author of Build a Large Language Model From Scratch. | 529.8K |
| @natolambert Nathan Lambert | Nathan Lambert: open model research and commentary (Interconnects). | 106.9K |
| @Tim_Dettmers Tim Dettmers | Creator of bitsandbytes (quantization); CMU professor and Ai2 researcher. | 50.0K |
DGX Spark and home-lab community
| Account | Why follow | Followers |
|---|---|---|
| @NVIDIAAI NVIDIA AI | NVIDIA's AI account: DGX Spark, GTC and open-model news. | 355.1K |
| @sudoingX Sudo su | Local LLM benchmarks on GPUs and DGX Spark clusters (e.g. DeepSeek V4.1 Flash across 2 Sparks). | 39.8K |
| @CryptoBandit100 CryptoBandit | Runs a multi-unit DGX Spark/GB10 home cluster and posts tok/s results (GLM 5.3 Flash vs DeepSeek V4.1 Flash). | 648 |
| @TheAhmadOsman Ahmad | Local AI and home GPU rigs; moderates r/LocalLLaMA. | 80.0K |
| @jun_song Jun Song | Open-source AI builder who posts about 1x/2x/3x DGX Spark setups. | 46.6K |
| @tysn_dev Ty | Home-lab local AI builder (4x GB10, 4x AMD MI50, 3090 Ti) posting candid takes on open-weight models and hardware prices. | 262 |
| @MiaAI_lab Mia | Runs frontier open models locally and shares serving recipes (e.g. GLM 5.3 Flash) that others use on single and dual DGX Sparks, plus hands-on model tests. | 42.6K |
Last updated Oct 24, 2026. Follower counts are from public X profiles on that date. Notes summarize what each account posts publicly. Not affiliated with any listed account. Suggest an account: hello@ragshelf.com.
The RAGShelf weekly
New local RAG tools, releases, and tested setups. One email a week, no spam.