OpenAI lists gpt-6.1-terra at $1 per million input tokens, half of Sol

A new model name, gpt-6.1-terra, has appeared on OpenAI’s API pricing page, listed at $1.00 per million input tokens and $5.00 per million output tokens at standard rates. That is half the price of gpt-6.1-sol.

The page shows prices only. At the time of writing, we found no blog post, release note or model documentation from OpenAI describing what Terra does, so this report is limited to what the table states.

Terra’s price across every processing tier

Terra sits in the flagship group with three other models: gpt-6-astra, gpt-6.1-sol and gpt-6-luna. On the standard tier, Terra costs $1.00 for input and $5.00 for output, with cached input at $0.05. Cache writes cost $1.25 for the 30-minute option and $2.50 for the 12-hour option.

Other tiers follow the same pattern. In the batch table, Terra is $0.50 input and $2.50 output, with cached input at $0.025. In the fast table, which the page says was renamed from priority processing on July 30, 2026, it is $2.00 input and $10.00 output, with cached input at $0.10.

One gap is visible. The Ultrafast table lists Astra and Sol but has no row for Terra, so the page does not show that speed tier for the new model. OpenAI said on October 8 that Ultrafast was rolling out for Sol in the API, Codex and ChatGPT Work.

The page does not give a context window, benchmark results, release date or modality support for Terra. It also does not say whether the model is available to every API account. Regional processing and FedRAMP endpoints carry a 10% uplift, the page notes, for models released on or after March 5, 2026, which would presumably include Terra, though the page does not spell that out for this model.

A middle rung that halves Sol’s bill

Within the flagship list, the standard input prices run $10.00 for Astra, $2.00 for Sol, $1.00 for Terra and $0.10 for Luna. Terra therefore takes the slot between Sol and Luna, at ten times Luna’s price and one tenth of Astra’s. Output prices keep the same 5-to-1 ratio to input across all four models and all three tiers shown.

The cached-input rate is the most notable line for heavy users. Terra’s $0.05 is 5% of its $1.00 input price. Our explainer on prompt caching notes that OpenAI cache reads on GPT-5.6 and later cost 0.1x the normal input rate, and 0.05x on GPT-6.1 Sol. Terra’s listing matches the Sol ratio, a 95% discount on repeated prompt text.

How the price compares with other APIs

Outside OpenAI’s lineup, Terra’s input price is in line with some rivals and well above others. Our coverage of Kimi K3.1 Pro, in the same prompt caching explainer, cites $2 per million input tokens through Vercel AI Gateway, twice Terra’s listed rate. At the cheap end, our DeepSeek API guide lists deepseek-flash at 1 to 2 yuan per million input tokens on a cache miss and 4 to 8 yuan for output, billed in a different currency and with peak and off-peak pricing, so a direct comparison with Terra’s flat dollar rates is not exact.

Price alone says little about quality. Without benchmarks or a description from OpenAI, it is not possible to say whether Terra is a smaller model, a tuned variant or a different product line. The name pattern, with Astra, Sol, Terra and Luna, suggests a family, but OpenAI has not described the ordering.

For developers, the practical point is limited for now. The model is listed with prices, but the page does not confirm how to call it or whether it is live for all accounts. Teams budgeting around Sol’s $2.00 and $10.00 rates can note that Terra, if its quality suits their workload, would cost half as much on standard, batch and fast processing alike. That is a conditional, not a claim OpenAI has made.

OpenAI has not yet commented on the listing, and we will update this report if documentation or an announcement appears.