Why We Cube IntelligenceEfficiency = Intelligence³ ÷ Price
- Low intelligence is useless. A cheap model that can't follow complex instructions, code, or reason is worthless regardless of price. Cubing the intelligence score heavily penalizes low-capability models — a score of 10 becomes 1,000 while a score of 60 becomes 216,000. Cheap but incompetent models don't falsely rank as "efficient."
- Capability scales exponentially. An increase in intelligence yields a disproportionate jump in real-world capability. A model scoring 60 isn't just 20% better than one scoring 50 — it can solve entirely new classes of problems (agentic planning, advanced logic, zero-shot coding) that the lower-scoring model simply cannot handle.
- Expensive ≠ better value. Price acts as a linear denominator. If a model is overpriced, its efficiency drops sharply. But if a model achieves breakthrough intelligence, the cubed numerator easily outpaces the price penalty. This isolates the sweet spot: maximum intelligence at fair cost.
| Int. Ranking |
Model |
Creator |
Price ($/1M Tokens) |
Intelligence |
Intelligence³ |
Efficiency Score |