GPT-6.1 Terra: Pricing, Specs and Sol vs Luna Comparison

On October 26, 2026, OpenAI added GPT-6.1 Terra (model ID gpt-6.1-terra) to its API. OpenAI describes it as "a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing" (OpenAI API changelog). This guide covers pricing, specs, how it compares with the rest of the GPT-6 lineup, and what switching from GPT-5.6 Terra actually costs, using only OpenAI's own documentation.

Short answer: half the price of 6.1 Sol, ten times GPT-6 Luna

GPT-6.1 Terra costs $1 per million input tokens and $5 per million output tokens (model page). That is exactly half of GPT-6.1 Sol ($2 / $10) and ten times GPT-6 Luna ($0.10 / $0.50), which makes it the third of the four GPT-6 models by price.

Compared with GPT-5.6 Terra ($2 / $12), input is 50% cheaper and output is about 58% cheaper. The context window (1,050,000 tokens) and max output (128,000 tokens) are unchanged, and the knowledge cutoff moves from February 16, 2026 to June 1, 2026.

As of October 27, 2026, OpenAI's public changelog and news feed do not mention availability in the ChatGPT app. What is confirmed is the API: Responses, Chat Completions and Batch.

GPT-6.1 Terra specs

OpenAI lists four example uses: matching product photos to catalog entries, interpreting or comparing technical drawings, acting on application screens, and drafting and revising text. In other words, it is aimed at image-heavy judgment work and screen automation.

One catch: tool calling requires the Responses API. If your app calls tools through Chat Completions, you will need to migrate those requests (OpenAI GPT-6 guide). Unlike Astra and 6.1 Sol, Terra supports none reasoning effort, which makes it a good fit for simple, latency-sensitive steps.

Pricing: GPT-6 lineup vs GPT-5.6 Terra

Standard prices per 1M tokens (prompts up to 272K tokens) from the OpenAI pricing page:

• Prompts over 272K input tokens are billed at 2x input/cache and 1.5x output for the whole request ($2 / $7.50 for Terra).
• Batch and Flex are 50% off ($0.50 / $2.50).
• Fast mode is 2x ($2 / $10), and Fast mode supports EU data residency for GPT-6.1 Terra.
• Cache writes cost $1.25 per 1M tokens for 30-minute retention and $2.50 for 12-hour retention. Regional processing adds 10%.

What switching from GPT-5.6 Terra saves per month

Three example workloads. These figures exclude cache-write charges and any change in reasoning tokens (reasoning tokens depend on effort level and task difficulty, and differ between models).

Formula: uncached input (M) x input price + cached input (M) x cached price + output (M) x output price. For case 3, GPT-6.1 Terra is 20 x $1 + 30 x $0.05 + 10 x $5 = $71.50. On list price alone, every pattern comes out at roughly half.

In practice, output quality matters more than the unit price. OpenAI itself recommends comparing GPT-5.6 Terra and GPT-6.1 Terra "on representative images and workflows using the same reasoning effort." If a cheaper model misses more defects in an inspection workflow, the human review cost eats the savings. Run around 100 real samples side by side before switching.

We also help companies connect models like this to their own data and build custom AI tools, starting from which model to choose. If that would be useful, feel free to take a look.

Terra vs Sol vs Luna vs Astra: which to use

OpenAI's guide positions the four models as follows: Astra for the most demanding reasoning, coding and professional work; 6.1 Sol for balanced speed, cost and intelligence ("near-Astra performance at a lower cost"); 6.1 Terra as fast and efficient for visual reasoning, computer use and natural writing; and Luna as the fastest and most cost-effective for focused, high-volume tasks.

This table is a starting point drawn from OpenAI's descriptions, not a benchmark. OpenAI has not published a numeric comparison between Terra and Sol, so the cost-sensible order is to start with Terra and move only the steps that fall short to Sol.

Migration checklist

1. Change the model name to gpt-6.1-terra.
2. Move tool calls to the Responses API if you use Chat Completions for tools.
3. Compare at the same reasoning effort (default medium), then raise it only if needed.
4. No audio or video input: text and image only.
5. No fine-tuning or predicted outputs: workflows that depend on them cannot move.

Summary

• GPT-6.1 Terra launched in the OpenAI API on October 26, 2026 (gpt-6.1-terra).
• $1 input / $5 output per 1M tokens: less than half of GPT-5.6 Terra, half of GPT-6.1 Sol.
• Built for visual reasoning, computer use and natural writing, with six reasoning effort levels from none to max.
• Switch after comparing real samples at the same effort level. ChatGPT availability is not confirmed.