OpenAI GPT 6.1 Terra API — Visual Reasoning & Writing: AI 大语言模型

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.

交互式模型控制

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.

📝

概览

关于此模型

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.

💰

价格与价值

成本分析

提供商	费用	备注
muapiapp	$1.00/M input, $0.05/M cached input, $5.00/M output	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 input, $0.05/M cached input, $5.00/M output

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.

** 竞品价格根据相似模型架构和使用层级估算。

⚙️

技术详情

配置参数

参数	类型	描述	默认值
提示词	string	用户消息或指令。	`Summarize this article in three bullet points.`
图像 URL	string	多模态请求的可选图像 URL。	`undefined`
系统提示词	string	可选的系统级指令，用于引导模型行为。	`You are an expert tutor.`
网页搜索开关	boolean	启用网页搜索工具以获取实时信息。	`false`
推理强度	枚举（6 个选项）	Reasoning depth: none, low, medium, high, xhigh, or max. Defaults to medium.	`medium`

提示词string

用户消息或指令。

默认值Summarize this article in three bullet points.

图像 URLstring

多模态请求的可选图像 URL。

默认值undefined

系统提示词string

可选的系统级指令，用于引导模型行为。

默认值You are an expert tutor.

网页搜索开关boolean

启用网页搜索工具以获取实时信息。

默认值false

推理强度枚举（6 个选项）

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

默认值medium

📖

实施指南

开发者文档

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.

❓

常见问题

常见问答

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.