GPT-6.1 Terra is designed for visual reasoning, precise computer interaction, and clear, natural writing. Start at the medium reasoning level, evaluate representative tasks, and increase the level only when the quality gains justify the added time and tokens.
Suitable tasks: Matching product photos to catalog entries, comparing technical drawings, operating application interfaces, and drafting or revising text.
Unsuitable tasks: The source does not list tasks that Terra is unsuitable for; do not infer that other tasks are unsupported.
Applicable model version: GPT-6.1 Terra (API model name: gpt-6.1-terra). The migration guidance mentions GPT-5.6 Terra but does not provide a more specific version number.
Applicable client, agent, or API: OpenAI API. The page recommends setting model in Responses API requests. If an application calls tools through Chat Completions, migrate those requests to the Responses API.
Recommended reasoning levels and parameters: Supports none, low, medium (default), high, xhigh, and max. Start with medium; increase the level only if the quality improvement is worth the additional time and tokens.
This source does not provide a complete, copyable prompt specific to GPT-6.1 Terra. The following workflow is organized from the model's official task positioning; it is not a prompt from the source:
Classify the request as visual analysis, interface operation, or writing and revision, and specify the desired deliverable.
For visual tasks, provide the images to compare and the matching criteria; for interface tasks, describe the target state and scope of operations; for writing tasks, provide the original text, audience, and revision goals.
Set model: "gpt-6.1-terra" in the API request and start with the medium reasoning level.
Evaluate quality and latency on representative samples. Increase the reasoning level only if the quality gains justify the additional time and tokens.
When migrating from GPT-5.6 Terra, compare representative images and workflows at the same reasoning level. If tool calls still use Chat Completions, migrate those requests to the Responses API.
Task type → Provide necessary inputs and success criteria → gpt-6.1-terra / medium
→ Evaluate on representative samples → Are the quality gains worth the extra time and tokens?
├─ Yes: Increase reasoning effort and retest
└─ No: Keep the current levelWhen comparing GPT-5.6 Terra with GPT-6.1 Terra, keep the reasoning level the same and use representative images and workflows. Use the Responses API for requests that call tools.
The page positions GPT-6.1 Terra as a fast, efficient model for visual reasoning, precise computer interaction, and clear, natural writing.
Tasks listed on the page include matching product photos to catalog entries, comparing technical drawings, operating application interfaces, and drafting and revising text.
Terra supports the none, low, medium (default), high, xhigh, and max reasoning levels. The official recommendation is to start with medium and increase the level if the quality improvement is worth the additional time and tokens.
For migration from GPT-5.6 Terra, the official recommendation is to compare the two at the same reasoning level using representative images and workflows. Requests that call tools through Chat Completions should be migrated to the Responses API.
The page's “Prompting best practices” section explicitly says these prompts are starting points for the GPT-6 family and that behaviors observed with GPT-6 Astra should be evaluated on the selected model and workload. This article therefore does not treat Astra-specific prompts from that section as verified recommendations for Terra.
The use cases and reasoning-level recommendations above come from OpenAI's official positioning of GPT-6.1 Terra. They are not independent benchmark results and do not guarantee the outcome or speed of any task.
The source provides no complete prompt specific to Terra, task evaluation data, or minimum effective reasoning level.
The “Prompting best practices” section addresses Astra behavior; its behavior descriptions and templates cannot be assumed to apply equally to Terra.
The source does not list task types that Terra is unsuitable for, nor does it provide quantitative quality or cost comparisons across models.
At collection time, the public Markdown version of the OpenAI documentation page (the original page URL with .md appended) was opened in a browser. The complete page was read, and the GPT-6.1 Terra section, reasoning-level guidance, migration advice, and scope of the prompting best practices were excerpted. The collection date was 2026-10-27; dynamic documentation may change later.
GPT-6.1 Terra