The value pick of the GPT-5.6 family — performance competitive with GPT-5.5 at roughly half the price. Built for the majority of everyday production work.
Terra lands at $2.50 / 1M input and $15 / 1M output — the same rate as the GPT-5.4 it replaces, and roughly 2× cheaper than Sol for competitive everyday quality.
Terra is the middle child of the GPT-5.6 family — the model OpenAI expects most developers to reach for day to day. Released publicly on July 9, 2026 alongside Sol and Luna, it holds the $2.50 / $15 per-1M rate of the GPT-5.4 model it replaces while matching much of GPT-5.5's quality at roughly half the cost.
Terra doesn't ship the max/ultra reasoning modes that make Sol distinctive; instead it targets the sweet spot of price and capability for high-volume production work. For teams migrating off GPT-5.5, it's the natural default — similar output quality, meaningfully lower spend.
Terra is a strong generalist: writing, code, vision, structured extraction, and reliable tool use across multi-turn loops. It gives up the deepest reasoning on the hardest problems (where Sol's Ultra mode earns its cost) but handles the vast majority of everyday tasks without the flagship price tag. The large context window keeps it viable for document Q&A and repo-scale coding.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| GPT-5.6 Terra | $2.50 | $15 | 1M |
| GPT-5.6 Sol | $5 | $30 | 1M |
| GPT-5.6 Luna | $1 | $6 | 1M |
| Claude Sonnet 5 | $3 | $15 | 1M |
| Gemini 3.1 Pro | $2 | $12 | 1M |
Terra's closest rivals are Claude Sonnet 5 (equal on output, slightly higher on input) and Gemini 3.1 Pro (cheaper, with a per-request tier surcharge above 200K tokens). Against its own family, it's the value option below Sol and above Luna.