AI vs Puzzles

Back to Home
Puzzle#1043

Connections

Released: 2026-04-19


7
Models Tested
1
Solved
0.86
Avg. Correct Connections
7
LLM Providers

google/gemini-2.0-flash-001

google

Failed

FLOP, BUST, DUD, HOLE

Misses

FRESH, MINT, ARCH, WISE

Words That Mean Sassy

CAP, KID, TURN, RIVER

Words That Can Precede "Over"

WAIST, LENGTH, HIPS, SASSY

Aspects of a Quilt


mistralai/mistral-medium-3

mistral

Failed

ARCH, BUST, FLOP, DUD

Types of failures

FRESH, SASSY, WISE, KID

Words that can precede ’guy’

HIPS, LENGTH, WAIST, HOLE

Measurements in clothing

RIVER, TURN, CAP, MINT

Words that can follow ’bank’


anthropic/claude-3.7-sonnet

anthropic

Failed

FRESH, SASSY, WISE, KID

DESCRIBING A PERSON’S ATTITUDE

ARCH, BUST, HIPS, WAIST

PARTS OF THE BODY

LENGTH, RIVER, TURN, CAP

SWIMMING POOL TERMS

FLOP, HOLE, DUD, MINT

CONDITIONS OF SOMETHING


openai/gpt-4o

openai

Failed

ARCH, BUST, CAP, FLOP

Words that can precede ’cap’ to make compound words

FRESH, SASSY, WISE, MINT

Adjectives describing a person’s demeanor or attitude

HIPS, WAIST, LENGTH, TURN

Measurements of the body

RIVER, HOLE, DUD, KID

Words related to things or beings


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

meta

1/4

MINT, FRESH, SASSY, WISE

Types of personality traits or descriptors

BUST, HIPS, WAIST, LENGTH

Measurements of the human body

FLOP, DUD, CAP, KID

Informal terms for failure, or types of hats, or colloquialisms for child

RIVER, HOLE, TURN, ARCH

Types of features or landmarks in geography or golf


deepseek/deepseek-chat-v3-0324

deepseek

1/4

ARCH, BUST, HIPS, WAIST

Parts of the body

FRESH, SASSY, WISE, KID

Words that can follow ’smart’

FLOP, HOLE, RIVER, TURN

Words that can follow ’belly’

CAP, DUD, LENGTH, MINT

Words that can follow ’pennant’


x-ai/grok-3-mini-beta

x-ai

4/4

BUST, HIPS, WAIST, LENGTH

Body measurements

FLOP, HOLE, RIVER, TURN

Poker terms

ARCH, FRESH, SASSY, WISE

Descriptive adjectives

CAP, DUD, KID, MINT

Common nouns


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