Own your intelligence, not a subscription.
AIVerseWeb maps the entire universe of AI you can run on your own hardware — models, tools and honest cost math — so you stop renting intelligence by the month.
live demo · runs on any modern laptop · no cloud, no account, no meter
39+
open models tracked
12
tools covered
$1,000+
possible yearly savings
Start here
Four doors into the AI universe
Whether you want to chat privately, code faster, generate images or understand the field — it all starts here.
Learn local AI
Why owning beats renting, what hardware you actually need, and your first model running in minutes.
View allFind your model
A curated directory of open-weight models filtered by what YOUR machine can run.
View allPick your stack
Ollama vs LM Studio vs llama.cpp — runtimes, UIs and workflows compared without hype.
View allCrunch the math
Subscriptions feel small. We total them up against one-time hardware and show your break-even date.
View allModel directory
40+ open-weight models, matched to real hardware
From 0.6B pocket assistants to 235B MoE giants — parameters, quants, VRAM fit and licenses at a glance.
Qwen3 8B
5.8GBThe default recommendation for a 12GB GPU: strong chat, solid code, hybrid thinking mode.
Qwen3 30B-A3B
19GBMixture-of-Experts trick: 30B knowledge with 3B speed. The people's flagship.
Llama 3.1 8B
5.8GBThe model that made local AI mainstream. Still a dependable daily driver.
DeepSeek-R1 Distill 32B
20GB24GB rigs' reasoning monster: shows its work, solves hard problems offline.
Gemma 3 12B
9GBVision + 128K context + great writing. A 12GB-card all-rounder.
Phi-4 14B
9.8GBMicrosoft's textbook-trained 14B — punches far above its weight in math and logic.
Fresh signal
Latest from the AIverse
Curated updates on models, hardware and tooling — summarized, dated, zero noise.
Digital product
The Local AI Playbook
Everything we know about building a private, subscription-free AI setup — hardware shopping lists, step-by-step stacks, prompt systems and maintenance routines.
✉ Newsletter
The Signal
One email per week: new models worth your RAM, tool updates and price math. No spam, unsubscribe anytime.