The meaningful difference
Both share the GPT-5.6 tool stack; Sol buys a higher capability ceiling while Terra lowers token costs by half.
Choose by workload
Choose GPT-5.6 Sol when your priority is the highest capability ceiling and expensive failures. Its relevant strengths include frontier agentic and coding results and million-token long-context evaluations.
Choose GPT-5.6 Terra when your priority is balanced everyday production work. Account for closed weights before committing.
Specification comparison
| Measure | GPT-5.6 Sol | GPT-5.6 Terra |
|---|---|---|
| Access | API + app | API + app |
| License | Proprietary | Proprietary |
| Context | 1.0M | 1.0M |
| Provider-listed API input / 1M | $5.00 | $2.50 |
| Provider-listed API output / 1M | $30.00 | $15.00 |
| MMLU-Pro | — | — |
| GPQA Diamond | 94.6 | 92.9 |
| SWE-bench Verified | — | — |
| LiveCodeBench | — | — |
| SWE-Bench Pro | 64.6 | 63.4 |
| Artificial Analysis Intelligence Index | 58.9 | 55 |
A fair test for this pair
Route the same production sample to both tiers and compare accepted outputs per dollar rather than list price alone.
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.