AI vs Puzzles

Back to Home
Puzzle#995

Connections

Released: 2026-03-02


7
Models Tested
5
Solved
3.43
Avg. Correct Connections
7
LLM Providers

openai/gpt-4o

openai

4/4

BEAM, HORSE, RINGS, VAULT

Gymnastics apparatus

POSITION, RANK, STANDING, STATION

Words associated with status or placement

BALL, FOUL, SAFE, STRIKE

Baseball terms

FUNKY, POPCORN, RUBBER, SPRING

Words associated with music or dance genres/styles


google/gemini-2.0-flash-001

google

2/4

RANK, STANDING, POSITION, STATION

Hierarchical Level

SPRING, RUBBER, BEAM, VAULT

Things That Bounce

FOUL, SAFE, STRIKE, BALL

Baseball Terms

HORSE, RINGS, POPCORN, FUNKY

Things That Can Be Done, Eaten, or Styled


anthropic/claude-3.7-sonnet

anthropic

4/4

BEAM, HORSE, RINGS, VAULT

GYMNASTIC EQUIPMENT

POSITION, RANK, STANDING, STATION

STATUS OR PLACE

BALL, FOUL, SAFE, STRIKE

BASEBALL TERMS

FUNKY, POPCORN, RUBBER, SPRING

TYPES OF CHICKEN


deepseek/deepseek-chat-v3-0324

deepseek

4/4

BEAM, HORSE, RINGS, VAULT

Gymnastics events

POSITION, RANK, STANDING, STATION

Terms related to status or hierarchy

BALL, FOUL, SAFE, STRIKE

Baseball terms

FUNKY, POPCORN, RUBBER, SPRING

Words that can follow ’Chicken’


x-ai/grok-3-mini-beta

x-ai

4/4

BEAM, HORSE, RINGS, VAULT

Gymnastics apparatus

POSITION, RANK, STANDING, STATION

Status terms

BALL, FOUL, SAFE, STRIKE

Baseball terms

FUNKY, POPCORN, RUBBER, SPRING

Styles of music


mistralai/mistral-medium-3

mistral

4/4

BEAM, HORSE, RINGS, VAULT

Gymnastics Events

POSITION, RANK, STANDING, STATION

Words meaning status

BALL, FOUL, SAFE, STRIKE

Baseball Terms

FUNKY, POPCORN, RUBBER, SPRING

Types of Music Genres


meta-llama/llama-3.3-70b-instruct

meta

2/4

BEAM, VAULT, RINGS, STANDING

Gymnastics apparatus

BALL, STRIKE, SAFE, FOUL

Baseball terms

HORSE, RANK, POSITION, STATION

Military terms

FUNKY, POPCORN, RUBBER, SPRING

Types of music or items with a spring


Methodology Note

Each model receives the same prompt with the 16 words to connect. Models are tasked with creating 4 connections and their answers are evaluated without feedback or retries.

Leaderboard