01 / Control Panel
Calculator Parameters
1.1x
0.5x (Optimistic)1.1x (Task Default)3.0x (Pessimistic)
Calculation Formula
Base cost is calculated per million tokens. The total outcome cost =
(in_tokens × $/M + out_tokens × $/M) × retry_rate
Loading interactive visualizations...
Loading interactive visualizations...
05 / Ground Evidence
Sorted by cost ascendingStack-Ranked Outcomes
| LLM Model | Estimated Tokens | Quality Score | Outcome Cost ▲ | Token Source |
|---|---|---|---|---|
| Gemini 3 FlashCheapest | In: 3,800 Out: 1,200 | 74% | $0.0007 | Aider Workload |
| GPT-4o mini | In: 3,900 Out: 1,400 | 76.5% | $0.0016 | Aider Workload |
| DeepSeek V4 Pro | In: 4,000 Out: 1,850 | 82.1% | $0.0037 | Aider Workload |
| Claude 4.5 Haiku | In: 3,800 Out: 1,300 | 79.1% | $0.0091 | Aider Workload |
| o3-mini | In: 4,500 Out: 2,500 | 85.6% | $0.0175 | Aider Workload |
| Gemini 3.5 Flash | In: 3,850 Out: 1,450 | 78.9% | $0.0207 | Aider Workload |
| GPT-4o | In: 4,100 Out: 1,900 | 84.8% | $0.0322 | Aider Workload |
| Gemini 3.1 Pro | In: 4,300 Out: 2,000 | 83.5% | $0.0359 | Aider Workload |
| Claude 4.6 Sonnet | In: 4,000 Out: 1,800 | 85.2% | $0.0429 | Aider Workload |
| Claude 4.7 Opus | In: 4,200 Out: 2,200 | 86.8% | $0.2508 | Aider Workload |
The Ground Truth Registry (⬇️ Sorted by True Cost)
What is measured: The raw underlying calculations proving the final costs. It maps the workload token counts, verified accuracy (Quality Score), and the calculated True Cost.
Verifiability & Dual Sourcing: (1) Quality Source: Click any highlighted link under Quality Score to audit the accuracy evaluation (e.g. LMSys battles or AgentNoah OWASP audits). (2) Token Source: The rightmost badge identifies the benchmark workload used to audit raw input/output token counts.