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

# GPT-6.1 Terra Is Live

> GPT-6.1 Terra is now listed on APIYI, with standard input at $1 and output at $5 per million tokens. Explore its visual reasoning focus, 1.05M context window, reasoning levels, API examples, and billing boundaries.

## Key points

* **Now listed on APIYI**: The model ID is `gpt-6.1-terra`. Standard input costs \$1, output \$5, and cache reads \$0.05 per million tokens, matching OpenAI's corresponding standard rates.
* **Visual and text tasks**: OpenAI positions the model for product image matching, technical drawing interpretation, screen interaction, and text revision. It accepts images and produces text.
* **1.05M context window**: The official context window is 1,050,000 tokens, with a maximum output of 128,000 tokens. Capacity limits and long-context billing need to be considered separately.
* **Reasoning can be disabled**: Available levels are `none`, `low`, `medium`, `high`, `xhigh`, and `max`. The default is `medium`.
* **Use Responses for tools**: The two API examples below cover ordinary text requests. Use `/v1/responses` when integrating function tools.

## Release background and positioning

OpenAI announced GPT-6.1 Terra in its API changelog on October 26, 2026. We checked the official documentation and APIYI's public pricing endpoints on October 27 and confirmed that the model has been added to the platform's model list, then completed the functional checks below.

Terra focuses on visual reasoning, computer interaction, and natural text expression. Developers can start by trying it on one specific step in an existing workflow: matching product images to catalog records, interpreting annotations in technical drawings, identifying interface states from screenshots, or revising an existing draft.

These are task areas described by OpenAI, and they still require validation against your business needs. Test screenshot accuracy, preservation of meaning in revised text, and tool argument reliability with your own samples. This article does not provide performance rankings or speed multipliers that we have not reproduced.

## Specifications and integration boundaries

| Item | Official specification |
| - | - |
| Model ID | `gpt-6.1-terra` |
| Context window | 1,050,000 tokens |
| Maximum output | 128,000 tokens |
| Knowledge cutoff | June 1, 2026 |
| Input / output | Text and images / text |
| Reasoning effort | `none`, `low`, `medium` (default), `high`, `xhigh`, `max` |
| Features | Streaming, function calling, structured outputs |

> [!NOTE]
>
> This table lists OpenAI's official specifications. APIYI's public catalog lists the `CodexResponses`, `default`, and `svip` groups; access depends on your API key's group. The tests below verify the current test key's interface behavior, not every group, tool, or the full context limit.

## API functional tests

On October 27, 2026 (UTC+8), we completed **50 generation requests** using one test key and up to 20 concurrent tasks to finish the checks faster. Every request was checked for its content and completion state, with no automatic retries. This was a functional test, not a capacity test.

| Check | Requests | Observed result |
| - | - | - |
| Preflight, English, Chinese, and arithmetic | 8 | Both APIs returned the expected results |
| Streaming | 4 | Correct text deltas and normal completion signals |
| JSON Schema | 4 | Fields, types, and specified values matched |
| Six reasoning levels and the default | 14 | Parameters accepted; simple questions answered correctly |
| Image understanding | 4 | Both APIs identified the specified colors in synthetic images |
| Responses tool round trip | 2 | Correct arguments, local tool result accepted, final answer 42 |
| Multi-turn conversation | 4 | Chat history and Responses `previous_response_id` retained the test information |
| Repeated cache requests | 6 | Correct text; cache observations differ below |
| Longer input checks | 4 | Both APIs retrieved the hidden marker at actual input lengths of 8,230 and 32,806 tokens |

**Cache behavior must be reported separately for each API.** Three consecutive requests reused the same prefix on each endpoint. Responses reported 2,653 cached tokens on the second and third requests; Chat reported no cache hit on all three. This run therefore confirms cache hits only for the Responses samples and does not establish cache savings for Chat.

The reasoning checks validate accepted parameters and simple-task results, not a quality ranking across effort levels. They do not establish that higher effort is always better. The run did not cover the million-token context limit, cache retention duration, actual billing above 272K input tokens, or long-term reliability.

## Parameter differences when migrating

GPT-6.1 Terra supports `none`. Do not apply GPT-6.1 Sol's restriction on disabling reasoning to Terra. When migrating an existing application, first keep the reasoning level, input samples, and output requirements consistent, then compare the results.

When reasoning effort is not `none`, follow the official migration guide and remove `temperature`, `top_p`, and `top_logprobs`; also omit `logprobs` from Chat Completions requests. For tool use, choose Responses, execute the tool with the arguments returned by the model, and send the result back.

## API examples

Install the OpenAI Python SDK with `pip install -U openai`. Set `APIYI_API_KEY` in your local environment first. Keep real keys out of scripts, frontend code, and commit history.

The basic parameter formats for both APIs were verified in this test run. These examples read the key from the environment. Start with a short text request before adding images, tools, or long documents.

```python Responses API theme={null}
import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["APIYI_API_KEY"],
    base_url="https://api.apiyi.com/v1",
)

response = client.responses.create(
    model="gpt-6.1-terra",
    reasoning={"effort": "low"},
    input="Make this sentence clearer without changing its meaning: Integration testing will begin after the API is confirmed.",
    max_output_tokens=1024,
)
print(response.output_text)
```

```python Chat Completions theme={null}
import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["APIYI_API_KEY"],
    base_url="https://api.apiyi.com/v1",
)

response = client.chat.completions.create(
    model="gpt-6.1-terra",
    reasoning_effort="none",
    messages=[{"role": "user", "content": "Explain in one sentence how to check whether an API response succeeded."}],
    max_completion_tokens=1024,
)
print(response.choices[0].message.content)
```

For full parameters and tool result handling, see [Responses API](https://docs.apiyi.com/en/api-capabilities/openai/native.md), [Compatible API](https://docs.apiyi.com/en/api-capabilities/openai/compatible.md), and [Response Handling](https://docs.apiyi.com/en/api-capabilities/openai/response-handling.md).

## Pricing and availability

### Standard pricing

The following prices were checked on October 27, 2026. Rates are in USD per million tokens, before recharge bonuses or account group discounts. They correspond to OpenAI's standard tier for requests with no more than 272K input tokens.

| Billing item | APIYI public price | OpenAI standard price |
| - | - | - |
| Input | \$1.00 | \$1.00 |
| Output | \$5.00 | \$5.00 |
| Cache reads | \$0.05 | \$0.05 |

> [!NOTE]
>
> Model prices align with the provider's official prices and may change when those prices change. This table is for reference; the **Model Pricing** tab in the top navigation is authoritative: [Model Pricing](https://docs.apiyi.com/en/models/index.md).

For example, 10,000 uncached input tokens plus 2,000 output tokens cost \$0.02 at the standard rates, excluding cache writes, additional tool fees, and group discounts. This is a billing example, not a fixed price per task.

### Caching and long context

APIYI's public configuration lists cache writes at \$1.25 per million tokens. OpenAI lists separate 30-minute and 12-hour retention periods at \$1.25 and \$2.50, respectively. The platform's public fields do not specify the retention period, so they do not establish that both modes are available.

> [!WARNING]
>
> OpenAI specifies that requests exceeding 272K input tokens are billed at twice the base input and cache rates and 1.5 times the base output rate for the entire request. APIYI's public configuration did not list Terra-specific tier conditions during this review. The official tier rates therefore should not be treated as verified platform charges, and the full context window should not be assumed to use base pricing throughout. Check billing in the console and run a small-cost test before submitting large requests.

The public catalog lists `CodexResponses`, `default`, and `svip`. Check the current console for model pricing, group access, and additional features. Recharge offers are covered in [Recharge Promotions](https://docs.apiyi.com/en/faq/recharge-promotions.md).

## Integration recommendations

Use a small, fixed set of samples to check three things: whether answers meet your requirements, whether the API parameters are compatible, and whether actual charges fit your budget. Include blurry images and screenshots with small text in visual tests; check facts and preservation of meaning in writing tests; verify arguments and result handoff in tool tests. Record the model ID, reasoning level, and token usage for each request so that differences can be reproduced.

## Sources

Sources checked on October 27, 2026 (UTC+8). External sources are shown as plain text for copying.

* OpenAI API changelog: `developers.openai.com/api/docs/changelog`
* OpenAI model specifications: `developers.openai.com/api/docs/models/gpt-6.1-terra`
* OpenAI standard pricing: `developers.openai.com/api/docs/pricing`
* OpenAI parameter and migration guide: `developers.openai.com/api/docs/guides/latest-model`
* APIYI live pricing: [Public Pricing API](https://api.apiyi.com/api/pricing), [Public Ratio Configuration](https://api.apiyi.com/api/ratio_config), and [Model Pricing](https://docs.apiyi.com/en/models/index.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.