The meaningful difference
Terra is the balanced tier; Luna cuts input and output list prices while giving up some peak reasoning performance.
Choose by workload
Choose GPT-5.6 Terra when your priority is stronger reasoning on mixed daily work. Its relevant strengths include balanced capability and cost and strong reported software-agent results.
Choose GPT-5.6 Luna when your priority is bounded high-volume tasks where latency and cost dominate. Account for lower peak reasoning than sol and terra before committing.
Specification comparison
| Measure | GPT-5.6 Terra | GPT-5.6 Luna |
|---|---|---|
| Access | API + app | API + app |
| License | Proprietary | Proprietary |
| Context | 1.0M | 1.0M |
| Provider-listed API input / 1M | $2.50 | $1.00 |
| Provider-listed API output / 1M | $15.00 | $6.00 |
| MMLU-Pro | — | — |
| GPQA Diamond | 92.9 | 92.3 |
| SWE-bench Verified | — | — |
| LiveCodeBench | — | — |
| SWE-Bench Pro | 63.4 | 62.7 |
| Artificial Analysis Intelligence Index | 55 | 51.2 |
A fair test for this pair
Use a stratified batch of routine and edge-case tasks to find where Luna's lower cost begins to increase retries or human correction.
Bottom line: use reported results to form a hypothesis, then make the decision with representative private tasks. Missing scores remain missing; no composite winner is manufactured.