gpt-6.2-astra API
GPT-6.2 Astra is the GPT-6.2 model OpenAI calls its most capable, aimed at the hardest projects such as building and revising software, computer use and document work, with text and image input and text output. Use the exact ID through XiuAI, call tools with Responses and assess costs from this page’s current prices.[1][2][4]
gpt-6.2-astraModel overview
Published provider specifications. Available features depend on the current API route.
- Input
- text / image
- Output
- Text
- Context
- 1.05M
- Maximum output
- 128000
- API routes
- OpenAI-compatible API / openai-response / openai-response-compact / Anthropic Messages / Gemini GenerateContent / openai-alpha-search
A model for the hardest tasks
OpenAI lists building and revising software, using a computer and working with documents, and its guide recommends Astra for the most demanding tasks. Computer use means the model can take part in such tasks; your application still provides the environment, permissions and result checks. XiuAI provides model access only.[1][2]
Choosing between Astra and GPT-6.2 Sol
The two model pages list the same context window, maximum output, modalities and endpoints. They differ in positioning, vendor list price (higher for Astra) and knowledge cutoff (May 18, 2026 for Astra, June 1, 2026 for Sol). Compare them on the same acceptance set and current XiuAI prices; “most capable” does not guarantee better results on your task.[1][3][4]
Default behavior and prompt changes
OpenAI’s guide notes that Astra writes more concisely and directly by default and may propose a next step and wait for confirmation. Unattended automation should state in its instructions whether to proceed without confirmation and how much detail to return, and should revisit existing “be concise” instructions.[2]
Parameter checks when moving to GPT-6.2
The GPT-6.2 API rejects a none reasoning setting. OpenAI suggests starting new integrations at medium and moving none or minimal settings to low; use reasoning.effort in Responses and reasoning_effort in Chat Completions. Omit temperature, top_p, logprobs and top_logprobs from Chat Completions requests. In Responses, omit temperature, top_p and top_logprobs, and remove message.output_text.logprobs from include.[1][2]
API access
Using GPT-6.2 Astra through XiuAI
Use gpt-6.2-astra with an existing authorized API key. Format text or image-understanding input with the current protocol example, and use Responses for function tools. If your application falls back to Sol or another model on errors, OpenAI treats a mid-conversation downgrade as a change and drops reasoning items, so validate the fallback path separately.[1][2][4]
Send a request using your existing API key.
POST /v1/chat/completionscurl 'https://router-api.xiu.ai/v1/chat/completions' \
-H "Authorization: Bearer $XIUAI_API_KEY" \
-H 'Content-Type: application/json' \
--data-raw '{
"model": "gpt-6.2-astra",
"messages": [
{
"role": "user",
"content": "Hello"
}
]
}'Read the API documentationFrequently asked questions
Why do requests fail with reasoning effort set to none or minimal?
The model page states that the API rejects none, and the guide recommends replacing none or minimal with low and starting new integrations at medium. Applications that only change the model ID while keeping an older reasoning setting must update that parameter too.[1][2]
Can I call tools through Chat Completions?
OpenAI states that both GPT-6.2 models support Chat Completions but require Responses for tool calling. A successful text request validates only that request; tool definitions, tool results and follow-up turns need separate validation with the Responses structure.[1][2]
Can it take audio or video input, or output images?
The model page lists text and image input with text output; audio and video are not supported. Image input does not mean image generation; use an appropriate available model for media generation or audio/video understanding.[1][4]
Why does earlier reasoning seem lost after a mid-conversation change?
Starting with GPT-6.2, the API drops reasoning items when it detects a mid-conversation change, including prefix edits, a model downgrade, or a different end user or organization. To keep reasoning continuity, leave the sent prefix unchanged and keep the same model within a conversation.[2]
Related models
gpt-6.2-solgpt-6-astragpt-6.1-solSources and updates
Specifications reference provider documentation; prices and availability come from the current XiuAI catalog. Specifications checked: 2026-11-18
Content reviewed: 2026-11-18
[1] OpenAI GPT-6.2 Astra documentation[2] OpenAI guide to using GPT-6.2[3] OpenAI GPT-6.2 Sol documentation[4] Current XiuAI model and pricing data