GENERALIST GAMING AGENTS / 2026

ONE AGENT.
MANY GAMES.

Can one agent pick up an unfamiliar real-time game, figure out the rules, and act at the right moment? We're building an open benchmark to find out.

INDEPENDENT RESEARCH INITIATIVE · PUBLIC COMPETITION IN PREPARATION

32REAL-TIME GAMES
10DIVERSE GENRES
16/16PUBLIC / PRIVATE LEADERBOARDS
$50KPRIZE POOL

01 / THE MISSION

THE NEXT LEVEL
IS GENERALIZATION.

Many game-playing agents are specialists. We test whether one system can adapt across unfamiliar real-time games, where timing matters as much as action selection.

A generalist must perceive a new interface, learn its rules, plan ahead, and time its actions as game state evolves, without custom code for each game. Real-time play brings perception, reasoning, and control together. Those same capabilities matter for computer-use agents and embodied systems.

SEE HOW WE EVALUATE
ACTION ✳ PUZZLE ✳ PLATFORMER ✳ STRATEGY ✳ SIMULATION ✳ RACING ✳ ROLE-PLAYING ✳ SHOOTER ✳ + MORE

02 / THE BENCHMARK

A WORLD OF GAMES.
ONE INTERFACE.

One offline model and one harness must handle the rules and timing of 32 real-time games. Four sample games support development; the other 28 are withheld.

01 — OBSERVE◉

SEE THE GAME

Read the game brief and track the action through the current screen and recent frames. Motion and changing state matter, not just a single screenshot.

02 — ACT⌘

TAKE CONTROL

Choose what to press, when to press it, and how long to hold it. Every game uses the same mouse-and-keyboard interface, with no game-specific adapter.

03 — MEASURE↗

MEASURE PROGRESS

Scores come directly from game state, normalized to 0–1—no subjective model judge.

One model + harness

16 public / 16 private leaderboard games

Offline, reproducible evaluation

Open research resources

03 / THE COMPETITION

READY, AGENT?
PRESS START.

Build one offline agent that plays 32 unfamiliar real-time browser games. The upcoming Kaggle competition has a $50,000 prize pool. Test adaptation, planning, and timing-sensitive control with one model and harness.

VIEW KAGGLE HOST PROFILE Competition launch details to be announced
KAGGLE RESEARCH COMPETITION
$50,000
TOTAL PRIZE POOL
32 GAMES1 AGENTOPEN METHODS
Retro pixel-art robot mascot holding a game controller

04 / RESEARCH DESIGN

MORE THAN
A LEADERBOARD.

Real-time games bring perception, planning, exploration, and control into one setting. An agent must choose not only what to do, but when and for how long. The goal is reproducible evidence of transfer across environments.

[01]

HELD-OUT GENERALIZATION

Four sample games are provided for development. The remaining 28 are withheld. Sixteen games contribute to the public leaderboard and sixteen to the private leaderboard.

+
[02]

TIMING-SENSITIVE CONTROL

The right action at the wrong moment can fail. Agents must track changing game state and coordinate when to press, hold, release, or switch inputs.

+
[03]

STATE-BASED SCORING

Each game exposes a programmatic performance signal instead of relying on screenshot interpretation or an AI judge.

+
[04]

OPEN, REPRODUCIBLE RUNS

The organizers are preparing shared harnesses, reference baselines, and pinned evaluation assets.

+
[05]

BEYOND THE COMPETITION

A larger organizer-operated game collection is planned for subsequent studies of cross-game transfer.

+

05 / THE PEOPLE

BUILT BY RESEARCHERS.

An independent, collaborative research effort. Organizers are listed in citation order.

Benedikt Schink

Benedikt Schink

University of Hamburg
Princeton University (visiting)

Matin Urdu

Matin Urdu

ETH Zürich
Princeton University (visiting)

Arian Urdu

Arian Urdu

University of Hamburg
Princeton University (visiting)

Wayne Chi

Wayne Chi

Carnegie Mellon University

Stephanie Milani

Stephanie Milani

Microsoft Research New York
New York University (visiting)

Chengshuai Shi

Chengshuai Shi

Princeton University

Fei Fang

Fei Fang

Carnegie Mellon University

Chi Jin

Chi Jin

Princeton University

Karthik Narasimhan

Karthik Narasimhan

Princeton University

Kiran Vodrahalli

Kiran Vodrahalli

Google DeepMind

Seth Karten

Seth Karten

Princeton University

06 / FAQ

QUICK
ANSWERS.

Does the agent have to play every game?

Yes. The central challenge is a single offline model-and-harness pair that handles the rules and timing of all 32 real-time games, rather than a different agent for each title.

Why is real-time gameplay a core challenge?

Game state changes over time. Agents need to track motion and choose when to press, hold, or release controls, not just select an action from a static screen. Real-time describes the gameplay; the final competition rules will specify evaluation timing and runtime limits.

What does the agent see and control?

It gets the game brief, objective, controls, and visual observations. It interacts through a universal mouse-and-keyboard interface.

Are the games and leaderboard public?

Four sample games are provided; the other 28 games are withheld during the competition. “Public” and “private” refer to the two 16-game leaderboard splits, not which game builds are distributed. Final rules and submission instructions will be available on Kaggle at launch.

When can I participate?

The competition is in preparation. Launch timing, submission instructions, and final rules will be announced on the Kaggle competition page.

What licenses will apply?

All 32 games have MIT-licensed code. We have requested MIT licensing for original competition resources and data, and for winning submissions’ original code. Third-party components, assets, and pretrained weights retain their applicable licenses. The final Kaggle rules will govern competition participation.

THE QUESTION REMAINS

CAN ONE AGENT
PLAY THEM ALL?

Help test generalization where perception, planning, and real-time control meet.

FIND US ON KAGGLE
Generalist Gaming Agents pixel-art robot mascot