★
Menu
Home How It's Built Watchlist The Stack
Hardware ChipsCompare Chips Servers & SystemsNetworking MemoryFoundries & Packaging Cooling
Facilities Data CentersPower ProjectsCloud & GPU Compute
Intelligence AI ModelsSoftware AI ProductsBenchmarks Research PapersDatasets Concepts
Ecosystem CompaniesStocks PeopleTimeline TrendsFuture of AI
Learn News Ask the Data AI Agents
About About us Methodology Data & API Disclaimer
Options
DATA & API

Free read-only JSON access to the site's cached data.

Dark mode

Guided View
New to chips, data centers and model jargon? We explain key terms as you browse, in plain English. Same data, with the help built in.

Expert View
You already know the stack. Just the data: clean, fast and compact, with no extra explanations. This is the default view.

Interface language
‹ Back
Home › AI Models › GPT-4

GPT-4

AI Model Language Historical

The March 2023 landmark that made AI a boardroom subject.

You're looking at an AI model record: who built it, what it can do, and the hardware and data behind it.

GPT-4 jumped capability far past GPT-3.5 — bar exams, competition math, nuanced writing — and its technical report disclosed neither parameters nor architecture, marking the industry's turn to secrecy. Community and analyst estimates (never confirmed) describe a large mixture-of-experts trained on ~25,000 A100s at ~2×10²⁵ FLOP. It set the bar every 2023–24 model chased.

Specifications & key facts

Each figure is labeled by basis: vendor spec, vendor claim, reported, estimate, or official. "Not publicly disclosed" means exactly that — we never invent values.

Overview

Parameters Not publicly disclosed MoE architecture per community estimates only
Training compute ~2×10²⁵ FLOPEstimate
Training hardware NVIDIA A100 fleet (Azure)Reported

How this connects to the rest of the stack

This record's links at a glance — click a group to jump to it.

Developed by

Trained on

Succeeded by

Successor to

Sources

GPT-4 technical report ↗

Record reviewed as of 2026-01-15 — see Methodology.