New VDF AI Enterprise is now available on AWS Marketplace
Insights · Research · Field Notes

Enterprise AI, explained for builders and buyers. Page 4 of 29

Practical writing on governed agent orchestration, on-premise AI, compliance, and the infrastructure decisions that separate pilot projects from production platforms.

Subscribe via RSS Explore white papers
261 Articles
39 Topics
12 Featured
Fibre optic cables plugged into a network switch in a server rack, a picture of the many system connections an AI agent reaches through its tools
AI Infrastructure 10 min read

Best Local LLM for Agents and Tool Calling (2026)

Which open-weight model to run behind a local AI agent in October 2026, judged on what agent loops need: well-formed function calls, the right tool across many turns, structured output, a context budget that survives long tool loops, and a serving stack that parses the model's tool-call format. Vendor benchmarks with their caveats, memory sizing for parallel agent runs, and parser settings for vLLM, llama.cpp and Ollama.

#tool calling#AI agents#local LLM
Read article
Rows of white archive boxes on wooden shelves in a records room, a picture of the document collections a retrieval-augmented generation system has to answer from
AI Infrastructure 10 min read

Best Local LLM for RAG (2026): Models That Stay Grounded

Which open-weight model to use as the generator in a private retrieval-augmented generation system, judged on what RAG needs: staying faithful to retrieved passages, reasoning across long contexts, following format instructions and citing sources. Hallucination and long-context scores from named sources, memory sizing for concurrent RAG prompts, and picks for each hardware tier.

#private RAG#RAG#local LLM
Read article
Top-down view of a Mac mini beside a keyboard and trackpad on a dark desk, the kind of compact Apple silicon machine people use to run local language models
AI Infrastructure 11 min read

Best Local LLM for Mac: Mac mini, MacBook Pro and Mac Studio (2026)

Which open-weight model to run on each Mac Apple sells in October 2026, from a 16 GB MacBook Air or Mac mini to a 512 GB Mac Studio. Memory tiers checked against Apple's spec pages, MLX 4-bit file sizes from Hugging Face, the context each model leaves, and how MLX compares with llama.cpp, Ollama and LM Studio.

#local LLM#Mac Studio#open-weight models
Read article
A laptop screen showing analytics dashboards with histograms and line charts, the kind of image a local vision language model can read and explain
AI Infrastructure 13 min read

Best Local Vision Models (2026): Open-Weight VLMs for Images, Charts and Screens

Open-weight vision language models you can run on your own GPUs, checked against their model cards in October 2026: sizes, licences including Llama 4's EU clause, memory for weights and image tokens, serving support in vLLM, llama.cpp and Ollama, and which model suits photos, charts, screens or video.

#vision language models#open-weight models#local LLM
Read article
A long aisle of library shelves, a reminder that a RAG framework is mostly about finding the right passage in a very large collection
RAG 12 min read

Best RAG Frameworks (2026): LlamaIndex, LangChain, Haystack and More

Open-source RAG frameworks checked against their GitHub repositories and documentation in October 2026: LlamaIndex, LangChain with LangGraph, Haystack, RAGFlow, DSPy, txtai, Kotaemon and LightRAG, compared on ingestion, retrieval, evaluation, agentic RAG and deployment, with the projects that are archived or in maintenance mode flagged.

#RAG#private RAG#RAG architecture
Read article
Close-up of a small controller chip surrounded by surface-mount components on a blue circuit board, a reminder that small language models are sized for modest hardware
Small Language Models 14 min read

Best Small Language Models (2026): SLMs Compared by Task

Eleven open-weight small language model families from 0.8B to 14B parameters, checked against their model cards in October 2026 and compared by task: tool calling, retrieval over long documents, image input and edge deployment, with licences, memory arithmetic and when a small model beats a large one.

#small language models#open-weight models#tool calling
Read article
A long aisle between rows of metal server racks, the kind of self-hosted infrastructure where a RAG team keeps its vector index next to the documents it was built from
RAG 15 min read

Best Vector Database for RAG (2026): Open-Source and Self-Hosted Options

Ten vector databases checked against their own documentation and GitHub repositories in October 2026: pgvector, Qdrant, Milvus, Weaviate, Chroma, OpenSearch, Elasticsearch, LanceDB, Redis and Vespa, plus Pinecone as the managed reference. Compared on licence, hybrid search, filtering, quantization, multitenancy and the work it takes to run each one yourself.

#vector database#vector search#private RAG
Read article
A triple-fan graphics card standing upright on a white surface, the kind of single GPU that the smaller Gemma 4 sizes are built to run on
AI Infrastructure 11 min read

Gemma 4 On-Premise: Local Requirements, Sizes and Setup (2026)

Google's Gemma 4 family runs from phone-sized E2B to a 31B dense model, all under Apache 2.0. This guide lists every size and variant, the memory each needs for weights and KV cache, Google's official quantization-aware builds, and the install steps for vLLM, llama.cpp, Ollama, LM Studio and MLX.

#Gemma#open-weight models#local LLM
Read article
Out-of-focus lines of colour-highlighted code on a dark monitor, standing in for the terminal commands that install and serve gpt-oss on local hardware
AI Infrastructure 10 min read

gpt-oss Local Deployment Guide: Requirements and Setup (2026)

OpenAI's gpt-oss-20b and gpt-oss-120b are Apache 2.0 reasoning models that ship in MXFP4, so they fit one 16 GB or one 80 GB device. This guide covers the memory each size needs at real context lengths, the harmony format and reasoning effort, tool calling, and step-by-step setup with vLLM, Ollama and llama.cpp.

#gpt-oss#open-weight models#local LLM
Read article
Stay ahead of enterprise AI

Turn insight into an on-prem AI roadmap.

Pair these articles with a product walkthrough to see how VDF AI handles orchestration, governance, and cost control inside your own infrastructure.

Book a demo Explore the platform