The living
scoreboard for AI.
Every capability is a race and every gap is real. See who leads, by how much and for how long, then learn to get the most out of whichever model you pick.
0
live races
language, image and video
0
models tracked
58 language models
0
photo finishes
leads inside the noise
0
lead changes
in the last 90 days
Every capability, its own track
Distance between racers is the real score gap, so a photo finish looks like one and a runaway lead looks like one too.
Dynasties, streaks and the labs that win most
Every week a model leads any race counts as a race-week at #1. Here is who has banked the most since January 2025.
Open leaderboardsRaces led right now, by lab
The front of the pack
Everything you need to know about each model in thirty seconds: what it is for, where it wins and how to prompt it.
Which agent actually finishes the job?
Agent rankings score complete jobs, from fixing a failing test suite to reconciling invoices, with the margin of every win shown plainly.
The same model, ten times better
Most bad answers are bad prompts. Learn the anatomy of a strong prompt, the words that change results, and what 8 official guides agree on.
Prompt upgrade
Before
write a blog post about remote work
After
From first prompt to production agent
Six paths, 32 lessons and 8 quizzes. Progress saves on this device, no account needed.
Your progress
0%
done
0
of 32 modules
0
of 8 quizzes passed
0
XP · Novice
Saved in this browser. No account needed.
Only what moves the race
Curated, impact-rated, and always with a plain explanation of why it matters.
Not sure where to start? Pick your lane.
I'm new to AI
Plain-English basics, safe habits and your first genuinely useful conversations.
Start AI FoundationsI'm choosing a model
Compare the leaders race by race and see what you pay for each point.
Open head-to-headI'm building with AI
Agent architectures, frameworks, evals and the failure modes to design out.
Explore agentsGroups tracked: Overall · Coding · Hard Reasoning · Math · Science · Writing & Language · Multimodal · Tool Use & Agents · Long Context · Speed & Cost · Trust & Factuality · Image Generation · Video Generation. Scores are illustrative placeholders. How scoring works