GPT 6.1 Terra API

Updated 29 Oct 2026

What is GPT 6.1 Terra?

GPT 6.1 Terra is OpenAI's mid-tier GPT-6.1 model, released on October 26, 2026 as the upgrade to GPT 5.6 Terra. It runs on UnificAlly as openai/gpt-6.1-terra. OpenAI built it for fast computer use, visual reasoning, and clear writing: reading what is on screen, acting across apps, and handling the forms, documents, and drawings that fill a workday. It sits between GPT 6 Luna and GPT 6.1 Sol and costs half as much per token as Sol. It takes text and image input, returns text, and works across a 1,050,000-token context window with a 128,000-token max output and knowledge up to June 1, 2026. Reasoning effort runs from none up to max.

What GPT 6.1 Terra can do

What GPT 6.1 Terra is best for

• Computer-use agents

  Form filling, data entry, and app workflows. 60.0% on Agents' Last Exam, level with GPT 6 Astra.

• High-volume document intake

  Reading invoices, receipts, and scans and entering the details into other software, at half GPT 6.1 Sol's token price.

• Visual reasoning

  Charts, screenshots, and technical drawings. 85.5% on MMMU-Pro at max effort, within 1.2 points of GPT 6 Astra.

• Writing and communication

  Proposals, reports, and technical explanations. OpenAI built Terra for clearer, more natural writing.

• Fast interactive replies

  Artificial Analysis measured about 1.7 seconds to first token at medium effort, against 3.6 seconds for GPT 6.1 Sol.

• Sampling control

  Set reasoning_effort: "none" to use temperature and top_p, which GPT 6.1 Sol does not allow.

GPT 6.1 Terra use cases

Put GPT 6.1 Terra behind a browser or desktop agent that works through back-office tasks: opening each incoming invoice, reading the totals, and entering them into accounting software. Send screenshots, scanned forms, and engineering drawings as image input and ask for the values or the next action. Draft proposals, summaries, and customer replies at low or medium effort, where it answers in about two seconds. Teams running agents on GPT 6.1 Sol or GPT 6 Astra can move the computer-use steps to Terra and keep the larger models for deep research and complex code changes.

GPT 6.1 Terra limitations

The default effort is much weaker than max. Artificial Analysis scores it 38 on its Intelligence Index at medium effort against 51 at max, so raise reasoning_effort for hard multi-step work.

Max effort uses a lot of tokens. Artificial Analysis counted about 55% more output tokens than GPT 6.1 Sol on the same test suite, which narrows Terra's cost advantage at that setting.

It is not the model for the hardest coding. GPT 6.1 Sol scores 55.9% on Terminal-Bench 4.0 and GPT 6 Astra 58.2%, against Terra's 50.3%, and OpenAI recommends them for the most complex coding tasks.

Long prompts cost more. Once a request's prompt passes 272,000 tokens, the whole request bills at the long-context rate: double on input, cache read, and cache write, and 1.5x on output.

GPT 6.1 Terra vs GPT 6.1 Sol

Terra costs half as much per token and answers faster. It matches or beats Sol on computer use and science: 60.0% against 58.6% on Agents' Last Exam and 58.3% against 57.0% on Terminal-Bench-Science 0.1, both at lower cost per task. Sol is better at coding (55.9% against 50.3% on Terminal-Bench 4.0) and desktop apps (78.0% against 74.2% on OSWorld 2.1). Pick Terra for agents that read screens and fill in forms, and Sol for agents that write and fix code.

When to use GPT 6.1 Terra

Use GPT 6.1 Terra as the default for computer-use agents, document intake, and everyday writing, where you want Astra-level results on screen-based tasks without paying for a larger model. Step up to GPT 6.1 Sol for coding agents and to GPT 6 Astra for the hardest research, and drop to GPT 6 Luna for high-volume classification and short replies.

GPT 6.1 Terra API examples

Call GPT 6.1 Terra from any language by POSTing to /v1/chat/completions, the OpenAI-compatible endpoint shared by every language model in the UnificAlly API. Full parameter docs live at docs.unifically.com/models/llm/openai/gpt-6.1-terra.

curl -X POST https://api.unifically.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "openai/gpt-6.1-terra",
    "reasoning_effort": "medium",
    "messages": [
      {
        "role": "user",
        "content": [
          { "type": "text", "text": "Read this invoice and return the vendor, date, and total as JSON." },
          { "type": "image_url", "image_url": { "url": "https://example.com/invoice.png" } }
        ]
      }
    ]
  }'

The response comes back synchronously with the completion. Set "stream": true to receive tokens as they generate, and raise reasoning_effort to high or max for long multi-step tasks.

GPT 6.1 Terra guides and comparisons

Review pricing, limits, and tested outputs before running this model.

GPT 6.1 Terra API pricing

OpenAI API pricing per 1M tokens for GPT 6 Astra, Sol, Luna and GPT 5: input, output, cache and long-context rates, real request costs, and which to pick.

Read GPT 6.1 Terra API pricing→