AISpendGuard

Cost vs Quality

Find the cheapest AI model that meets your quality bar. Each point represents a model's average cost per call plotted against its quality score from Lab benchmarks. The dashed line marks the efficiency frontier — models that offer the best quality at each price point.

$0.0000$0.0003$0.0006$0.0009$0.001$0.00145%56%67%78%89%100%Average Cost per Call (USD)Quality ScoreClaude Haiku 4.5: $0.0003/call, 83% quality, 4 runsClaude Sonnet 4.6: $0.001/call, 50% quality, 4 runsGPT-4o: $0.0004/call, 100% quality, 4 runsGPT-4o Mini: $0.0000/call, 100% quality, 4 runs
AnthropicOpenaiEfficiency frontier

How to read this chart

  • Top-left = best value (high quality, low cost)
  • Dashed line = efficiency frontier — models on this line are Pareto-optimal (no other model is both cheaper and better)
  • Quality scores are based on human verdicts: Correct = 100%, Partial = 50%, Incorrect = 0%
  • Models with fewer than 3 verdicts are excluded to avoid misleading results