GPT-5.5 Pro 73·GPT 46·GPT-5.4 Pro 45·GPT-5.2 Pro 30·GPT-5.2 Thinking 29·GPT-5 24·GPT-5.4 Thinking 19·GPT-5.5 15·GPT-5.4 10·GPT-6 Astra 10·GPT-5.6 Sol 7·GPT-5.2 6·GPT-5.5 Thinking 5·GPT Pro 4·GPT-5.6 4·GPT-5.2 multi-agent system 1·GPT-5.3 Instant 1
Systems
73 AI systems named in 650 claims, grouped by who makes them and by product line, each with the record computed from its claims. These are counts of recorded work, not a ranking of ability: which problems attract attention has little to do with how hard they are.
Every system
OpenAI
Codex 30·GPT-5.2-Codex 1
Harmonic
Google DeepMind
Gemini Deep Research 14·Gemini 8·Gemini 3 Pro 8·Gemini 3.1 Pro 8·Gemini 3 Flash 6·Gemini Pro 3·Gemini 3 2·Gemini Deep Think 2·Gemini 3.1 1·Gemini 3.1 Deep Think 1·Gemini Flash 1
Anthropic
Claude 24·Claude Opus 4.7 9·Claude Opus 4.6 8·Claude Fable 5 4·Claude Opus 4.5 4·Claude Opus 4.8 4·Claude Mythos 2·Claude Opus 2·Claude Sonnet 4.6 2·Claude Opus 5 1·Claude Sonnet 4.5 1
Axiom Math
ByteDance Seed
Logical Intelligence
FrenzyMath
Project Numina
UlamAI
Wes Pegden, Carnegie Mellon University
Archivara
DeepSeek
ETH Zürich SRI Lab
Math Inc
Principia Math
Shouqiao Wang and Davide Crapis
xAI
Maker not recorded
Each maker is attributed from a source, linked on each system's page. Claims whose source names no system, or several without saying which, appear on their problems but are not counted here.