> ## Documentation Index
> Fetch the complete documentation index at: https://docs.anyfast.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# GPT-6 Astra

> Use OpenAI GPT-6 Astra through the AnyFast Responses API for complex reasoning, coding, research, computer use, and document workflows.

GPT-6 Astra is OpenAI's GPT-6 flagship model for complex reasoning, coding, research, computer use, and document creation. Use the model ID `gpt-6-astra` with the OpenAI-compatible Responses API at `POST /v1/responses`.

<Info>
  `gpt-6-astra` is available from both `POST /v1/responses` and `POST /v1/chat/completions`. Use the Responses API for tools, image and file input, Structured Outputs, and prompt caching.
</Info>

## Model specifications

| Property         | Value                                   |
| ---------------- | --------------------------------------- |
| Model ID         | `gpt-6-astra`                           |
| Recommended API  | Responses API                           |
| Input            | Text and image                          |
| Output           | Text                                    |
| Context window   | 1,050,000 tokens                        |
| Maximum input    | 922,000 tokens                          |
| Maximum output   | 128,000 tokens                          |
| Knowledge cutoff | April 30, 2026                          |
| Reasoning effort | `low`, `medium`, `high`, `xhigh`, `max` |
| Fine-tuning      | Not supported                           |

## Key capabilities

* **Long-context reasoning** — Analyze large codebases, documents, and multi-step workflows within a 1.05M-token context window.
* **Software engineering** — Plan and execute long-running coding tasks across files, tools, and verification steps.
* **Vision** — Accept image input together with text instructions.
* **Structured Outputs** — Return JSON that follows a supplied schema.
* **Prompt caching** — `prompt_cache_key` and `prompt_cache_options` support the `30m` cache lifetime and expose cache-write and cache-hit token counts.
* **Responses API tools** — Web search, code interpreter, hosted shell, apply patch, computer use, MCP, function tools, and custom tools are available through AnyFast.
* **Async tool calling** — Function and custom tools accept `async: true` and return the corresponding tool-call item.

## Capability examples

<Tabs sync={false}>
  <Tab title="Text and reasoning">
    Use `reasoning.effort` to balance latency and reasoning depth. GPT-6 Astra accepts `low`, `medium`, `high`, `xhigh`, and `max`.

    ```bash cURL theme={null}
    curl https://www.anyfast.ai/v1/responses \
      -H "Authorization: Bearer YOUR_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "model": "gpt-6-astra",
        "input": "Review this retry design, identify race conditions, and propose a safe implementation plan.",
        "reasoning": {
          "effort": "high",
          "summary": "auto"
        },
        "max_output_tokens": 4096
      }'
    ```
  </Tab>

  <Tab title="Image and file input">
    Add an `input_image` part for diagrams, screenshots, or charts. Use `input_file` with inline Base64 data when you need to analyze a document.

    ```bash cURL theme={null}
    curl https://www.anyfast.ai/v1/responses \
      -H "Authorization: Bearer YOUR_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "model": "gpt-6-astra",
        "input": [{
          "role": "user",
          "content": [
            { "type": "input_text", "text": "Find the failure point in this architecture diagram." },
            { "type": "input_image", "image_url": "data:image/png;base64,<BASE64_IMAGE_DATA>", "detail": "high" },
            { "type": "input_file", "filename": "deployment-notes.pdf", "file_data": "data:application/pdf;base64,<BASE64_PDF_DATA>" }
          ]
        }],
        "reasoning": { "effort": "high" }
      }'
    ```
  </Tab>

  <Tab title="Structured output">
    Set `text.format.type` to `json_schema` and provide a strict schema when your application needs machine-readable output.

    ```bash cURL theme={null}
    curl https://www.anyfast.ai/v1/responses \
      -H "Authorization: Bearer YOUR_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "model": "gpt-6-astra",
        "input": "Classify the deployment risk as low, medium, or high.",
        "text": {
          "format": {
            "type": "json_schema",
            "name": "deployment_risk",
            "strict": true,
            "schema": {
              "type": "object",
              "properties": {
                "risk": { "type": "string", "enum": ["low", "medium", "high"] },
                "reason": { "type": "string" }
              },
              "required": ["risk", "reason"],
              "additionalProperties": false
            }
          }
        }
      }'
    ```
  </Tab>

  <Tab title="Tool workflows">
    Use the Responses API for function calling. OpenAI's GPT-6 Astra guidance does not support tool calling through Chat Completions.

    ```bash cURL theme={null}
    curl https://www.anyfast.ai/v1/responses \
      -H "Authorization: Bearer YOUR_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "model": "gpt-6-astra",
        "input": "Check the deployment status for service api-gateway.",
        "tools": [{
          "type": "function",
          "name": "get_deployment_status",
          "description": "Return the current deployment status for a service.",
          "parameters": {
            "type": "object",
            "properties": {
              "service": { "type": "string" }
            },
            "required": ["service"],
            "additionalProperties": false
          },
          "strict": true
        }],
        "tool_choice": "auto"
      }'
    ```
  </Tab>
</Tabs>

## Reasoning configuration

| Effort   | Use case                                          |
| -------- | ------------------------------------------------- |
| `low`    | Lower-latency requests and routine tasks          |
| `medium` | General reasoning and coding                      |
| `high`   | Complex analysis and multi-step work              |
| `xhigh`  | Difficult professional and engineering tasks      |
| `max`    | Maximum reasoning depth for the hardest workflows |

<Warning>
  GPT-6 Astra does not support `none` or `minimal`. It also does not support `temperature`, `top_p`, or `top_logprobs`. Do not send `logprobs` with Chat Completions.
</Warning>

## API guidance

* Use the Responses API for tool calling.
* For explicit prompt caching, use `prompt_cache_options.ttl` rather than the older `prompt_cache_retention` field.
* Use `include: ["reasoning.encrypted_content"]` with `store: false` when you need encrypted reasoning items for stateless workflows.
* Use `max_tool_calls` to cap built-in tool calls and `parallel_tool_calls` to control parallel function calls.
* Public image URLs must permit server-side downloads. Data URLs are supported when the remote host blocks automated access.
* Image input is supported; audio and video input are not supported.

<Warning>
  AnyFast does not currently expose the Responses retrieval, Conversations, or Vector Stores management endpoints. Do not depend on `background`, `conversation`, `previous_response_id`, stored prompt references, or `configuration_update` for this model. Cross-request item references can also fail when requests are routed to different upstream resources.
</Warning>

## Core parameters

| Parameter                 | Type            | Required | Description                                                                    |
| ------------------------- | --------------- | -------- | ------------------------------------------------------------------------------ |
| `model`                   | string          | Yes      | Must be `gpt-6-astra`                                                          |
| `input`                   | string / array  | Yes      | Text, image, file, or tool-result input                                        |
| `instructions`            | string          | No       | System-level instructions for this response                                    |
| `reasoning.effort`        | string          | No       | `low`, `medium`, `high`, `xhigh`, or `max`                                     |
| `reasoning.summary`       | string          | No       | `auto`, `concise`, or `detailed`                                               |
| `reasoning.context`       | string          | No       | `auto`, `current_turn`, or `all_turns`                                         |
| `reasoning.mode`          | string          | No       | Use `standard`; Pro mode is not exposed on the current AnyFast route           |
| `max_output_tokens`       | integer         | No       | Maximum generated tokens, up to 128,000                                        |
| `max_tool_calls`          | integer         | No       | Maximum total built-in tool calls processed by the response                    |
| `text.format`             | object          | No       | Text or Structured Outputs configuration                                       |
| `text.verbosity`          | string          | No       | `low`, `medium`, or `high`                                                     |
| `tools`                   | array           | No       | Function or Responses API tool definitions                                     |
| `tool_choice`             | string / object | No       | Controls whether and which tool is called                                      |
| `parallel_tool_calls`     | boolean         | No       | Allows multiple tool calls in one model turn                                   |
| `include`                 | array           | No       | Requests supported supplementary response data                                 |
| `prompt_cache_key`        | string          | No       | Improves cache matching for related requests                                   |
| `prompt_cache_options`    | object          | No       | Configures implicit or explicit prompt caching; `ttl` currently supports `30m` |
| `prompt_cache_breakpoint` | object          | No       | Marks an explicit reusable-prefix boundary inside a content block              |
| `stream`                  | boolean         | No       | Stream response events over SSE                                                |
| `stream_options`          | object          | No       | Controls stream-specific behavior such as event obfuscation                    |
| `store`                   | boolean         | No       | Controls upstream response storage; AnyFast does not expose response retrieval |
| `metadata`                | object          | No       | Application-defined string key-value pairs                                     |
| `truncation`              | string          | No       | Use `disabled` on the current AnyFast route                                    |

<Card title="API Reference" icon="code" href="/api-reference/model-api/openai/gpt-6-astra">
  View the GPT-6 Astra Responses API reference.
</Card>

## Official references

* [GPT-6 Astra model page](https://developers.openai.com/api/docs/models/gpt-6-astra)
* [Using GPT-6 Astra](https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-astra)

<script src="/public/feedback.js" />