Runnable examples for OpenAI GPT-6.2 Astra and GPT-6.2 Sol through PoYo's OpenAI-compatible Chat Completions and Responses endpoints: plain requests, reasoning effort, streaming, and the official OpenAI SDKs with a changed base URL.
Answer: To use the GPT-6.2 API, send a server-side request to POST https://api.poyo.ai/v1/chat/completions or POST https://api.poyo.ai/v1/responses with "model": "gpt-6.2-astra" or "model": "gpt-6.2-sol" and a PoYo API key. With the OpenAI SDK, set the base URL to https://api.poyo.ai/v1. The examples in this repository run with Python 3.9+, Node.js 22+, or cURL.
Last verified against the PoYo API reference and pricing on November 18, 2026.
- Model IDs:
gpt-6.2-astra(frontier tier) andgpt-6.2-sol(efficient tier) - Released: November 17, 2026 by OpenAI
- Endpoints: Chat Completions (
/v1/chat/completions) and Responses (/v1/responses), both with streaming - Context: 1,050,000 tokens, up to 128,000 output tokens per response
- Input: text and images; output is text
- Reasoning effort:
low,medium(default),high,xhigh,max, set withreasoning.effortin Responses - Price on PoYo: Astra $6.40 input / $32.00 output, Sol $1.60 input / $8.00 output per 1M tokens, 20% below OpenAI's standard rates
- Create a key in the PoYo dashboard.
- Copy
.env.exampleto.envand replace the placeholder. - Run one of the examples. Each prints the model's text and the token usage.
POYO_API_KEY=YOUR_POYO_API_KEY_HERE
POYO_BASE_URL=https://api.poyo.aipython python/generate.py # Chat Completions, gpt-6.2-sol
python python/generate.py payloads/responses.json # Responses, gpt-6.2-astra, high effort
python python/generate.py payloads/responses-stream.json # Responses streaming
python python/generate.py payloads/chat-completions.json gpt-6.2-astra # same request, other modelpython/generate.py uses only the Python standard library. Payloads with messages go to Chat Completions and payloads with input go to Responses.
node node/index.mjs
node node/index.mjs payloads/chat-completions-stream.json
node node/index.mjs payloads/responses.json gpt-6.2-solnode/index.mjs has no dependencies.
curl -X POST "https://api.poyo.ai/v1/chat/completions" \
-H "Authorization: Bearer $POYO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "gpt-6.2-sol", "messages": [{"role": "user", "content": "Explain the CAP theorem in three sentences."}]}'Responses, reasoning effort, and streaming requests are in curl/generate.md.
Existing OpenAI clients only need a new API key and base URL.
import os
from openai import OpenAI
client = OpenAI(api_key=os.environ["POYO_API_KEY"], base_url="https://api.poyo.ai/v1")
response = client.responses.create(
model="gpt-6.2-sol",
input="When should a web service return HTTP 409 instead of 400?",
reasoning={"effort": "low"},
)
print(response.output_text)import OpenAI from "openai";
const client = new OpenAI({ apiKey: process.env.POYO_API_KEY, baseURL: "https://api.poyo.ai/v1" });
const completion = await client.chat.completions.create({
model: "gpt-6.2-astra",
messages: [{ role: "user", content: "Write a one-line commit message for: fix off-by-one error in pagination." }],
});
console.log(completion.choices[0].message.content);Runnable versions with Chat Completions, Responses, and streaming: python/openai_sdk.py (pip install openai) and node/openai-sdk.mjs (cd node && npm install).
| File | Endpoint | Model | Task |
|---|---|---|---|
chat-completions.json |
Chat Completions | gpt-6.2-sol |
Find and fix a pagination bug |
chat-completions-stream.json |
Chat Completions, streaming | gpt-6.2-sol |
Three-line release note |
responses.json |
Responses | gpt-6.2-astra |
Zero-downtime UUID migration plan, reasoning.effort: high |
responses-stream.json |
Responses, streaming | gpt-6.2-sol |
Optimistic vs pessimistic locking, reasoning.effort: low |
Pass a model ID as the second argument to the Python or Node.js example to run any payload on the other model.
| Field | Requirement | Meaning |
|---|---|---|
| model | required | gpt-6.2-astra or gpt-6.2-sol. |
| messages | required | system, user, and assistant messages with content. |
| max_tokens | optional | Maximum generated tokens. |
| temperature | optional | 0–2. |
| top_p | optional | 0–1. Use either top_p or temperature. |
| stream | optional | true returns server-sent events. |
| Field | Requirement | Meaning |
|---|---|---|
| model | required | gpt-6.2-astra or gpt-6.2-sol. |
| input | required | A string, or role / content items with input_text and input_image blocks. |
| instructions | optional | System or developer instructions. |
| reasoning | optional | {"effort": "low"} through {"effort": "max"}. |
| max_output_tokens | optional | Maximum output tokens. |
| stream | optional | true returns server-sent events. |
| store | optional | Whether to store the response. |
| previous_response_id | optional | Continue a conversation from an earlier response. |
Both endpoints return OpenAI's native response shape, without the code / data wrapper used by PoYo generation tasks. Chat Completions text is in choices[0].message.content. Responses text is in the output_text blocks of output[] items with type: "message", and the OpenAI SDKs expose it as output_text.
Billing is per token. 1 credit = $0.005.
| Model | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
gpt-6.2-astra |
1,280 credits ($6.40) | 6,400 credits ($32.00) | 64 credits ($0.32) | 1,600 credits ($8.00) |
| OpenAI standard rate | $8.00 | $40.00 | $0.40 | $10.00 |
gpt-6.2-sol |
320 credits ($1.60) | 1,600 credits ($8.00) | 16 credits ($0.08) | 400 credits ($2.00) |
| OpenAI standard rate | $2.00 | $10.00 | $0.10 | $2.50 |
All prices are per 1M tokens. Example: a request with 10,000 input tokens and 2,000 output tokens costs $0.032 (6.4 credits) on Sol and $0.128 (25.6 credits) on Astra. Reasoning tokens are part of output_tokens and bill as output. Confirm current prices on the GPT-6.2 Astra and GPT-6.2 Sol model pages.
| GPT-6.2 Sol | GPT-6.2 Astra | |
|---|---|---|
| OpenAI positioning | Go-to model for everyday work: coding, clear writing, multistep agents | Frontier model for the hardest projects: professional documents, science, complex coding |
| PoYo price per 1M tokens | $1.60 input / $8.00 output | $6.40 input / $32.00 output |
| Context and output | 1,050,000 / 128,000 tokens | 1,050,000 / 128,000 tokens |
| Start here for | High-volume coding, agent steps, chat, drafting | Final deliverables, research analysis, large refactors |
- Default to Sol. It costs a quarter of Astra per token.
- Escalate to Astra for slides, spreadsheets, and memos that must follow your templates and your field's conventions, for scientific analysis run step by step, and for complex changes in large codebases, where a failed run costs more than the token difference.
- Raise effort before switching models. Try Sol at
highorxhighbefore moving a task to Astra, and keepmediumas the starting point for new integrations. - Route in code. Both models take the same request on the same endpoints, so routing is a change to
model: pick by task type, or run Sol first and retry on Astra when a check fails (tests, schema validation, a reviewer step).
- Chat Completions: each
data:line is achat.completion.chunk. Text arrives inchoices[0].delta.content, a chunk near the end carriesusage, and the stream ends withdata: [DONE]. - Responses: each
data:line is an event whosetypenames it. Concatenateresponse.output_text.deltaevents;response.completedcarries the final response andusage. Ignore event types you do not handle.
Treat a stream that closes before data: [DONE] (or, for Responses, before response.completed) as failed and retry it. The examples in this repository exit with an error in that case.
Set an application timeout long enough for high reasoning effort, retry only 429 and transient 5xx responses, and record usage from every response, including input_tokens_details.cached_tokens and output_tokens_details.reasoning_tokens. Do not put POYO_API_KEY in browser or mobile client code. Check the HTTP status and the error object rather than a code field. More in docs/production-notes.md.
How much does the GPT-6.2 API cost? On PoYo, GPT-6.2 Astra costs $6.40 per 1M input tokens and $32.00 per 1M output tokens; GPT-6.2 Sol costs $1.60 and $8.00. Both are 20% below OpenAI's standard rates. Cache reads cost $0.32 (Astra) and $0.08 (Sol) per 1M tokens.
What is the difference between GPT-6.2 Astra and GPT-6.2 Sol? Astra is OpenAI's frontier GPT-6.2 model for the hardest professional, scientific, and coding work. Sol is the efficient model for everyday coding, writing, and agents at a quarter of Astra's token price. Both share the 1,050,000-token context window and 128,000-token output limit.
Can I use the OpenAI SDK? Yes. Set the base URL to https://api.poyo.ai/v1, use your PoYo API key, and set model to gpt-6.2-astra or gpt-6.2-sol.
Should I use Chat Completions or Responses? Both work with both models. Use Chat Completions to drop GPT-6.2 into existing chat code. Use Responses for reasoning.effort, instructions, previous_response_id, and structured image input.
Does it support streaming? Yes. Set "stream": true on either endpoint.
What are the context window and max output? 1,050,000 tokens of context and up to 128,000 output tokens per response, for both models.
payloads/— Chat Completions and Responses request bodiespython/generate.py— standard-library Python example, streaming includedpython/openai_sdk.py— OpenAI Python SDK examplenode/index.mjs— dependency-free Node.js example, streaming includednode/openai-sdk.mjs— OpenAI Node.js SDK examplecurl/generate.md— copy-paste cURL requests, including streamingdocs/production-notes.md— limits and rollout checklist
- GPT-5.6-API — GPT-5.6 Sol, Terra, and Luna through the Responses API
- GPT-5.5-API — GPT-5.5 examples
- Gemini-4-Carbon-API — Gemini 4 Carbon in Chat Completions and Gemini Native formats
- awesome-poyo-api — every PoYo model repository in one list
- poyo-devtools — PoYo CLI, MCP bridge, and Agent Skill
- GPT-6.2 Astra API on PoYo
- GPT-6.2 Sol API on PoYo
- OpenAI models on PoYo: API reference
- Chat Completions API reference
- Responses API reference
- PoYo model catalog
MIT