Decision Arena.
You make the call. Then compare recorded model answers, measured request times, and reported costs.
Recorded results · 56 questions · 1,198 successful model responses · Collected 2026-10-27. Answers are replayed from real OpenRouter requests.
- ✓
- 2Pick your answer
- 3
Your callEvaluation · Score
How good is the agent's reply?
Customer: “Your API returns 429 errors constantly.” Agent reply: “You're hitting the rate limit of 100 req/min. Here's how to batch requests, or I can raise your limit.”
Pick your answer · or press 1–4
6 models answered this question
Their answers stay hidden until you pick yours.
ModelAnswer · ProbabilitySpeedCost
Jev 1.14TypeSafeTypeSafe? hidden––$0.000015
Decider V1.2PerplexityPerplexity? hidden––$0.0000027
GPT-6 Luna DecisionsOpenAIOpenAI? hidden––$0.000016
d1-ultraLiquid AILiquid AI? hidden––$0.0000037
ClefCloudflareCloudflare? hidden––$0.0000086
Microsoft-Decision-1MicrosoftMicrosoft? hidden––$0.0000034
Probability is the model’s probability for its selected answer. Cost is the charge reported by OpenRouter; — means unavailable. Speed replays the measured request time, including network and routing.
Your model match
Answer 10 questions to find the model that decides like you.
Every answer counts. After 10, we rank all 23 models by how often they agreed with you, with their speed and cost.
0/10