AI infographics
Live charts that update with the data: which model for which task, what runs on your hardware, free and local alternatives. Embed any of them on your site.
- Which AI model for which task — 2026 For every task: the strongest model, the best value for money and the best open-weights model — from the leaderboards with prices, updated daily.
- Paid AI → free, open source and local alternatives What the paid AI apps and coding assistants cost, what their free tier gives, and what replaces them for free: open source projects and apps that run models on your own computer.
- Which open model fits your hardware — 2026 Open-weights models against popular graphics cards and computers: the recommended quantization (the largest up to Q8 that still gives 10+ tokens per second, otherwise a Q4-class one; below Q3 only when nothing else fits), where it runs (graphics memory, unified memory or offloaded to RAM) and an approximate generation speed.
- Types of AI models: a map and a glossary LLM or VLM, open weights or open source, base, instruct, reasoning, distilled, MoE, GGUF and quantization — what the words in model names mean, with live examples.
- PC builds for local AI models, 20B to 120B — 2026 From one 16 GB graphics card to three RTX 3090s, an EPYC server and unified-memory machines: what each build runs and how fast, with the source of every measurement.
- AI models by year: who released what Every notable model release month by month: vendor and its country, open or closed weights, type, context window, size of open models — and which tasks the models of each year lead today.
- Free limits of AI services — 2026 What the free tier of each chatbot, coding agent, API and automation service actually gives, what it is best for, and when we last checked the limit.
- Map of AI tools: top 3 in every category — 2026 Chat apps, coding agents, local model runners, APIs, automation and servers — the three tools to start with in each category, with pricing and GitHub stars.
- fedi-index: AI model rating on open data Our own rating of language models on open-licence data: human votes and independent benchmarks on one scale.
Two questions behind the charts
These infographics answer the questions people ask us most often about AI, and they are drawn from the same data as our leaderboards and tool cards. When a rating, a price or a model file changes, the chart changes with it, so there is no frozen image to go stale. They fall into two groups.
- Choosing a model or a service. Which AI model for which task picks three models per leaderboard, AI models by year lists who released what, and paid AI and its alternatives sets ChatGPT-style apps and coding assistants against free, open-source and local options. Free limits of AI services shows what each free tier really gives, and the map of AI tools names three tools to start with in every category.
- Running an open model yourself. Types of AI models decodes the words in model names, which model fits your hardware checks memory and speed, and PC builds for local AI shows complete machines from 20B to 120B.
Where the numbers come from
Ratings are taken from LMArena, prices from models.dev and LiteLLM. Parameter counts, architecture and the sizes of quantized files come from Hugging Face; memory bandwidth from the hardware vendors' specifications; measured speeds from public benchmark reports, each with its link. Every figure states the date of its data underneath, and the speed of a local model is an estimate unless it is marked as measured.
Embed, download, reuse
Each chart has an "Embed on your site" button. It builds a short snippet that loads the figure in an iframe through embed.js: you choose the language (any of the site's languages), a light, dark or automatic theme, the height and the columns to keep. The embedded copy refreshes by itself, so a blog post written today still shows current data next year.
If you would rather draw the chart yourself, the same rows are published as JSON at /api/ai/infographics/{key}.json, with field names, sources and the update date. Our own data — curation, hardware, builds, the glossary and the fit estimates — is released under CC BY 4.0: you may reuse it freely, commercially too, as long as a link to fedi.software stays next to it. Third-party data keeps the licence of its source (LMArena CC BY 4.0, models.dev and LiteLLM MIT), and every figure lists those sources with their notices.
Editions by year
The model chooser, the timeline, the hardware matrix and the builds keep a yearly edition. Switching to a past year shows which models were new then and which have since dropped out — a quick way to see how fast this field moves.