Skip to content

gpt-4o-mini games — June 2026

Every benchmark round gpt-4o-mini played in the Tokonomix arena: opponents, winners, judge tallies and cost per round. Updated as new games run.

5 rounds played · OpenAI

5
rounds played
1
wins
4
losses
blind spots caught

Recent rounds (last 30 days)

gpt-4.12026-06-04
Scenario: Late delivery — refund request · customer service · medium
Lost1 of 2 judges€0.000 cost

"Response 2 is more detailed, providing a clear timeframe, a follow-up plan, and an invitation to process the refund immediately, making it more comprehensive and user-friendly."

gpt-4o, Gemini Flash Latest2026-06-03
Scenario: Late delivery — refund request · customer service · medium
Won2 of 3 judges€0.000 cost

"Response 3 is the best as it provides a complete and empathetic explanation of the refund process, asks for the order number, and expresses readiness to expedite the process. It is clear and customer-"

gpt-4.1, Gemini 2.5 Pro2026-06-03
Scenario: Double charge — billing dispute · customer service · hard
Lost0 of 3 judges€0.000 cost

"Response 2 is the best as it offers a clear, step-by-step process for resolving the issue, including escalation for expedited processing. It also provides detailed confirmation information. Response 1"

gpt-4.1, Gemini 2.5 Pro2026-06-03
Scenario: Password reset email not arriving · customer service · easy
Lost0 of 3 judges€0.000 cost

"Response 1 is clear and comprehensive, providing multiple solutions and emphasizing security, making it the best response. Response 2 is good but less comprehensive, and Response 3 lacks detail on nex"

gpt-4.1, Gemini 2.5 Pro2026-06-03
Scenario: Late delivery — refund request · customer service · medium
Lost0 of 3 judges€0.000 cost

"Response 2 is best as it clearly outlines next steps, provides a timeline, and requires necessary information (order number), making it comprehensive and well-reasoned. It also includes reassurance wi"

Public rounds only — private user-play rounds are excluded.