gpt-6.1-terra
OpenAI GPT-6.1 Terra supports text and image input with text output. It provides a 1,050,000-token context window and up to 128,000 output tokens. Use Responses for tool calling; Chat Completions is also supported.
Modalities
In / Out Price
$1/ 1M
$1 Official price
$5/ 1M
$5 Official price
Context
1.05M
Released
Oct 25, 2026
API call examples
POST /v1/chat/completionscurl --request POST 'https://api.tokatlas.com/v1/chat/completions' \
--header "Authorization: Bearer $UNIROUTER_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6.1-terra",
"messages": [
{
"role": "user",
"content": "Explain quantum entanglement."
}
],
"temperature": 0.7
}'Providers & routes
Current provider · OpenAI
Platform configured price · Group: codex
USD / 1M tokens. Each row shows input and output from the same tier. Displayed rates use the lowest available group multiplier; your API key group may differ.
GPT-6.1 Terra for code and interface review
Suited to tasks that mix code, screenshots, and written requirements. Provide acceptance criteria so the result can be checked against tests or visual evidence.
Cross-file context
Trace the impact of changes across components, callers, and tests.
Visual issue tracing
Map screenshot discrepancies to components and acceptance criteria.
Configurable reasoning
Evaluate effort separately for routine analysis and multi-step investigations.
Room for migration plans
Include affected paths, compatibility requirements, and tests in migration results.
About
Vision, code, and workflows
Terra handles code, document text, and interface screenshots in the same task, with native text and image input and text output. Its 1,050,000-token context and 128,000-token output cap support work across many specifications.
From visual evidence to validation
Give it screenshots and expected behavior to identify visible differences, component locations, and steps to check. An application that exposes UI tools can validate targets and arguments before recording the result.
Effort settings and comparison with Sol
Terra exposes six efforts from none to max. Establish separate baselines for simple classification and multi-step review. When comparing it with Sol, record time to an accepted result and retries.
Integration checks
Upstream function tools use the Responses API. Test images, streaming events, and response structures on the chosen endpoint. Include fields required by existing consumers in compatibility tests.
Computer use depends on the application's screenshots and available tools. Verify both the control selected and the resulting screen state. For visual interpretation alone, use the same screenshot to compare localization, reading errors, and supporting evidence.
gpt-6.1-terra FAQ
Can Terra skip reasoning on simple tasks?
Use none on routes that expose it. For classification or field extraction, test it first and compare latency and accuracy with higher effort.
Can Terra analyze screenshots directly?
Pair the screenshot with a written question to get a text response. Point out the region to inspect and check API Usage for the image request format.
How do I avoid truncation when sending a large data set?
The published limits are 1.05 million tokens of context and 128,000 generated tokens. Estimate the input first, set an output limit and budget for reasoning within it.
Which API should I use for tool calls?
OpenAI exposes native tool calls through Responses. When moving from Chat Completions, confirm the route and test both the tool arguments and the result format.