OpenAI GPT 6.1 Terra API — Visual Reasoning & Writing: AI大規模言語モデル

Use GPT 6.1 Terra through Muapi for visual reasoning, coding, and writing. Text and image inputs, six reasoning levels, web search, and SSE streaming.

インタラクティブなモデルコントロール

GPT 6.1 Terra provides efficient visual reasoning, image analysis, coding, and clear writing. Supports text and image input, optional web search, and reasoning effort from none to max. Standard pricing: $1.00/M input, $0.05/M cached input, $5.00/M output. Long-context rates apply above 272K input tokens.

📝

概要

このモデルについて

GPT 6.1 Terra is OpenAI's efficient model for visual reasoning and natural writing. Use Muapi's prompt-based API for image-grounded analysis, document drafting, coding assistance, and extraction tasks. It accepts a text prompt, optional image_url and system_prompt, six reasoning levels, and optional web search. Submit a prediction and poll for the result, or read text as an SSE stream. Standard token pricing is $1.00/M uncached input, $0.05/M cached input, and $5.00/M output, with a $0.0002 minimum. Prompts above 272,000 input tokens use 2x input/cache rates and 1.5x output rates.

1Visual Review: Interpret product images, screenshots, charts, and technical drawings.

2Writing: Draft proposals, reports, summaries, and customer communications.

3Coding: Explain code, diagnose bugs, and propose changes.

4Research: Enable web_search_switch when current information is needed.

💰

料金と価値

コスト分析

プロバイダー	費用	備考
muapiapp	$1.00/M 入力, $0.05/M cached 入力, $5.00/M 出力	Minimum $0.0002 per call. Above 272K input: 2x input/cache rates and 1.5x output rates. Cache writes $1.25/M at standard context.

muapiapp$1.00/M 入力, $0.05/M cached 入力, $5.00/M 出力

Minimum $0.0002 per call. Above 272K input: 2x input/cache rates and 1.5x output rates. Cache writes $1.25/M at standard context.

** 競合サービスの料金は類似のモデル構成および利用ティアに基づいて算出された推定値です。

⚙️

技術詳細

設定スキーマ

パラメータ	型	説明	デフォルト
プロンプト	string	生成プロセスを適切に制御するためのパラメータ設定（The user message or instruction.）。	`Summarize this article in three bullet points.`
画像 URL	string	処理の開始点となる入力元画像のURLまたはBase64データ。	`undefined`
System プロンプト	string	生成プロセスを適切に制御するためのパラメータ設定（Optional system-level instruction to guide model b）。	`You are an expert tutor.`
Web Search Switch	boolean	出力ファイルの保存形式（例：mp4、png、jpg、webpなど）。	`false`
Reasoning Effort	Enum（6個の選択肢）	Reasoning depth: none, low, medium, high, xhigh, or max. Defaults to medium.	`medium`

プロンプトstring

生成プロセスを適切に制御するためのパラメータ設定（The user message or instruction.）。

デフォルト値Summarize this article in three bullet points.

画像 URLstring

処理の開始点となる入力元画像のURLまたはBase64データ。

デフォルト値undefined

System プロンプトstring

生成プロセスを適切に制御するためのパラメータ設定（Optional system-level instruction to guide model b）。

デフォルト値You are an expert tutor.

Web Search Switchboolean

出力ファイルの保存形式（例：mp4、png、jpg、webpなど）。

デフォルト値false

Reasoning EffortEnum（6個の選択肢）

Reasoning depth: none, low, medium, high, xhigh, or max. Defaults to medium.

デフォルト値medium

📖

実装ガイド

開発者ドキュメント

Submit a prediction

POST /api/v1/gpt-6-1-terra with a required prompt. Add image_url for visual analysis and system_prompt to guide tone or behavior. reasoning_effort accepts none, low, medium (default), high, xhigh, or max. Set web_search_switch to true for web search.

curl -X POST "https://api.muapi.ai/api/v1/gpt-6-1-terra" \
  -H "x-api-key: $MUAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"Draft a concise proposal for automating invoice review.","reasoning_effort":"medium"}'

Poll /api/v1/predictions/{request_id}/result using the returned request_id. An optional webhook_url receives the result for polling requests.

Streaming

POST the same prompt fields to /api/v1/gpt-6-1-terra/stream. Read data: {"choices":[{"delta":{"content":"..."}}]} chunks until data: [DONE]. Streaming does not deliver webhooks.

These are single-turn text-generation endpoints. They do not expose raw Responses input items, custom function tools, file inputs, or structured-output schemas.

❓

よくある質問

FAQ

How much does GPT 6.1 Terra cost?

Standard rates are $1.00/M uncached input, $0.05/M cached input, and $5.00/M output, with a $0.0002 minimum per request. Cache-write tokens, when reported, cost $1.25/M. Above 272,000 input tokens, input and cache rates double and output costs $7.50/M.

Which reasoning levels are supported?

Use reasoning_effort: none, low, medium, high, xhigh, or max. The default is medium. More reasoning can increase latency and output token usage.

Can Terra analyze images?

Yes. Send image_url alongside prompt. The endpoint generates text; it does not generate images.

Can I stream responses?

Yes. POST /api/v1/gpt-6-1-terra/stream for live text deltas over SSE. The standard endpoint returns a request_id to poll.

Can Terra search the web?

Set web_search_switch to true to enable web search. Custom function calling, raw Responses items, and structured-output schemas are not exposed by these prompt-based endpoints.

What is the context window?

The model supports a 1,050,000-token context window, including up to 922,000 input tokens and 128,000 output tokens. The prompt API does not expose an output-token limit control.