Intelligence benchmarks
AA Intelligence Index
Illustrative snapshot inspired by Artificial Analysis Intelligence Index cost-per-task (Answer / Reasoning / Cache write / Cache hit / Input). Totals match published headline figures where noted; segment splits are approximate for charting. Not live OpenCodex metering.
| Model | Effort | Score | Cost / task | Score / $ | Use case |
|---|---|---|---|---|---|
| claude-fable-5 | max | 60 | $3.25 | 18.5 | Planner, Frontier |
| gpt-5.6-sol | max | 59 | $1.04 | 56.7 | Frontier, Planner |
| gpt-5.6-sol | xhigh | 58 | $0.88 | 65.9 | Frontier, Workhorse |
| claude-opus-4.8 | max | 57 | $1.78 | 32.0 | Frontier, Planner |
| gpt-5.5 | xhigh | 56 | $0.99 | 56.6 | Frontier, Workhorse |
| gpt-5.6-terra | max | 55 | $0.55 | 100.0 | Workhorse |
| gpt-5.6-luna | max | 51 | $0.21 | 242.9 | Workhorse, Cheap subagent |
| gemini-3.1-pro | preview | 46 | $0.32 | 143.8 | Frontier, Fast |
| deepseek-v4-proBest value | max | 44 | $0.04 | 1100.0 | Cheap subagent, Workhorse |
| grok-4.3 | high | 48 | $0.28 | 171.4 | Workhorse, Fast |
| glm-5.2 | max | 42 | $0.12 | 350.0 | Workhorse, Cheap subagent |
| claude-sonnet-4.6 | max | 49 | $1.15 | 42.6 | Workhorse |