NodeShift Blog

Featured blog post

June 23, 2025

How to Install Nano-VLLM Locally?

Nano-vLLM is a stripped-down, no-fluff engine designed purely for blazing-fast offline inference with large language models. It’s lightweight (just ~1,200 lines of code) but packs a serious punch — featuring smart optimizations like prefix caching, tensor parallelism, CUDA graphs, and more. Whether you’re testing models locally or building a custom inference stack, Nano-vLLM gives you raw speed, full transparency, and zero dependency bloat. It mirrors the vLLM API for easy migration, while staying small enough to dive into and hack on. If you’re running models like Qwen3-0.6B on your own GPU or a cloud VM — this is your toolkit.

All blog posts

June 21, 2025

How to Install and Run Mistral Small 3.2 24B

Just like earlier, a new Mistral’s Mistral-small model update has come, where neither the model is small nor this new update. In fact, it has got such new features that make it a modern polished model to be used by developers and enterprises. Mistral Small 3.2 24B is here. It tightens up where it counts, better instruction following, cleaner outputs, and a noticeable drop in those frustrating infinite loops. The numbers speak for themselves, Arena Hard v2 performance has more than doubled over the last version, and it now handles code generation, STEM reasoning, function calling and even vision tasks with surprising precision for its size. If you’re building assistants, agents, or apps that need dependable responses, this model brings a welcome balance of capability and control. It’s fast, focused, and frankly, not “small” at all.

June 20, 2025

How to Install Virtuoso Large 72b Locally?

Virtuoso-Large (72B) is a powerful open-source language model built on the Qwen2.5-72B foundation. Designed to handle complex and domain-spanning tasks, this model stands out for its depth, adaptability, and clarity. Whether you’re summarizing reports, writing technical content, generating multilingual insights, or diving deep into diverse datasets—Virtuoso-Large delivers high precision and natural language understanding at scale. Released under the Apache-2.0 license, it’s freely available for both personal and commercial use, making it a solid choice for anyone looking to build real-world applications without restrictions.

June 19, 2025

Guide to Install Kimi-Dev 72B: The Most Powerful Open-Source Coding LLM

If you’re working on software engineering tasks and tired of code generation models that break at the first test, meet Kimi-Dev-72B, a coding LLM that actually understands how real-world development works. This open-source model has just set a new benchmark on SWE-bench Verified, scoring 60.4% and outperforming all its open-source peers. Kimi-Dev is trained using large-scale reinforcement learning inside Docker containers, where it only gets rewarded if the entire test suite passes. That means fewer hallucinations, more robust patches, and code you can actually trust in production. If you’re building tools, fixing bugs, or automating dev workflows, Kimi-Dev brings serious engineering muscle to the table, and it’s free to use and extend.

June 18, 2025

How to Install Fanar-1 9B Arabic-English LLM Locally?

Fanar-1-9B-Instruct is a bilingual Arabic-English large language model developed by Qatar Computing Research Institute (QCRI) at HBKU, fine-tuned from Google’s Gemma-2-9B. Trained on 1 trillion tokens with a strong focus on Arabic dialects (Gulf, Levantine, Egyptian) and Modern Standard Arabic (MSA), it’s designed for culturally aware conversations, government/civic applications, and educational tools. With 4.5M instructions and 250K DPO preference pairs, it’s aligned with Islamic values and excels in multilingual Q&A, dialogue, and cultural understanding.

June 17, 2025

How to Install and Run MonkeyOCR Locally

In a landscape crowded with document AI tools, MonkeyOCR is taking its spot as a solution designed for serious document parsing tasks, especially when accuracy, structure, and speed matter. It is built on a novel Structure-Recognition-Relation (SRR) triplet paradigm, and redefines document understanding by replacing clunky multi-stage pipelines with a simple, unified model that achieves both performance and efficiency. It performs very well across diverse document types, from complex formulas and dense tables to multi-page text-heavy PDFs, in both Chinese and English. With a lightweight 3B parameter count, it outperforms heavyweight competitors like Gemini 2.5 Pro and Qwen2.5 VL-72B on average English document performance, while being significantly faster than models like MinerU and Qwen2.5 VL-7B in real-world multi-page scenarios. MonkeyOCR also shines with a remarkable +15% gain in formula parsing and +8.6% in table accuracy compared to modular systems, making it an essential tool for researchers, businesses, and developers handling structured documents at scale.

June 16, 2025

How to Install Jan-Nano Locally?

Jan-Nano is a lightweight yet powerful 4-billion parameter language model tailored for deep, research-heavy workloads. Built by Menlo Research, this model isn’t just about generating text — it’s designed to think clearly, answer precisely, and connect well with research tools thanks to its MCP (Model Context Protocol) compatibility. It’s fine-tuned from Qwen3-4B, which means it benefits from a solid foundation, but Jan-Nano takes it a step further with performance tweaks specifically for structured queries and fact-first tasks.

June 14, 2025

How to Install V-JEPA 2 by Meta: Enable Real-World Interaction in Robots & AI Agents

Meta’s latest breakthrough in world models, V-JEPA 2 (Video Joint Embedding Predictive Architecture 2), is here to expand the abilities of visual understanding and physical-world prediction. It is trained on over a million hours of video and powered by Meta’s JEPA architecture, and as a 1.2B-parameter world model, this model excels in recognizing objects, predicting future frames, and enabling zero-shot robot planning, without requiring prior exposure to specific environments. With benchmarks like Something-Something v2 (SSV2) and Epic-Kitchens-100 under its belt, V-JEPA 2 is a powerful system capable of simulating the consequences of hypothetical actions and planning complex sequences in new surroundings. This makes it a game-changer for anyone building AI agents, robotics applications, or video-based understanding systems. If you’re a researcher, ML engineer, or into Robotics, understanding how to get this model running locally or on the cloud is the first step toward building the next generation of intelligent, real-physical-world-interacting AI.

June 13, 2025

Transform Documents and Images into Markdown & HTML with Nanonets-OCR-s

Optical Character Recognition (OCR) has come a long way from simply pulling raw text out of images, and Nanonets-OCR-s is the perfect proof. This cutting-edge image-to-markdown OCR model doesn’t just extract content, it transforms documents into semantically rich, structured markdown that’s ready for downstream use by Large Language Models (LLMs). If you’re working with academic papers, legal contracts, business reports, or scanned forms, Nanonets-OCR-s ensures your image or PDF data is perfectly readable even in HTML/markdown formats and it’s contextually understood. With features like LaTeX equation recognition (handling both inline and block-level math), intelligent image description using structured tags, and powerful table extraction in markdown/HTML formats, this model offers precision and structure at an unmatched level. It even handles special elements like checkboxes, watermarks, and signatures with semantic tagging that makes them machine-friendly and human-readable.

June 12, 2025

How to Install and Run MiniCPM 4.0 8B Locally

MiniCPM4-8B is not just another large language model, it’s a cutting-edge breakthrough that brings flagship-level performance to your local machine with outstanding efficiency. With 8 billion parameters trained on 8 trillion tokens, this model delivers top-tier natural language understanding and generation while being developed specifically for blazing-fast inference, even on edge devices. What sets MiniCPM4-8B apart is its meticulously optimized architecture that features InfLLM v2, a sparse attention mechanism that processes long text (up to 128K tokens) with just 5% of the usual computation. It’s further backed by the UltraClean and UltraChat v2 datasets, ensuring high-quality pretraining and fine-tuning, and powered by CPM.cu and ArkInfer for rapid, cross-platform deployment. If you’re building chatbots, research tools, or AI agents, MiniCPM4-8B gives you the ability of using a heavyweight LLM with the agility of an optimized lightweight system.

June 11, 2025

How to Install Mistral Magistral Locally?

Magistral-Small-2506 is the latest evolution in Mistral AI’s line of efficient reasoning models, fine-tuned from the base Mistral-Small-3.1-2503. While it retains its compact size and agility, this model steps into deeper waters—bringing long-form reasoning, step-by-step deduction, and multilingual support into a package that runs comfortably on consumer-grade GPUs. What makes Magistral stand out is its clarity of thought. The model doesn’t just answer questions—it takes time to think. It writes out its reasoning process like a person solving a math problem on paper, making it especially useful for logic, science, code, and educational tasks. Whether you’re building a chatbot that needs to explain itself or a backend service for research-style outputs, Magistral-Small brings structure and depth with minimal overhead.

Trusted by Thousands of Cloud Professionals

Ready to build
with us?

The ideal way for organizations young and old to ease their way into the distributed and affordable cloud at their own pace.

Stay Tuned!

Stay up to date with the latest updates, news, and hotfixes for our product.

NodeShift creates a vital link between developers and affordable cloud.

Switch theme

English (EN)
Arabic (AR)
Chinese (ZH-CN)
German (DE)
Korean (KO)
Russian (RU)
French (FR)
Spanish (ES)
Portuguese (PT)
Japanese (JA)

JavaScript is disabled in your browser. For a better experience, please enable JavaScript.Learn how to enable JavaScript.