GPT-6 Astra API Guide: Pricing, Model ID, Context & Migration
Use GPT-6 Astra through the OpenAI API with the correct model ID. Check pricing, the 1.05M-token context window, supported tools, reasoning settings, and migration changes before production rollout.
Verification: Model IDs, pricing, availability, specifications, and code patterns were checked against the OpenAI sources listed below. Get Model Key has not independently run GPT-6 Astra through the API, so the examples should be verified in your own project before production use.
API access is billed separately from ChatGPT subscriptions. GPT-6 Astra is not supported on the API Free tier.
What is GPT-6 Astra and when was it released?
OpenAI announced GPT-6 Astra on September 3, 2026. It is the company's flagship model for difficult end-to-end work across reasoning, coding, research, computer use, and document creation. The API model ID is gpt-6-astra.
Availability is rolling out in phases across ChatGPT plans and the API. An API key does not guarantee immediate model access: your organization, usage tier, region, and rollout status still apply. Check the model list returned to your project before treating access as production-ready.
- Release date: September 3, 2026
- API model ID: gpt-6-astra
- Best fit: complex, multi-step professional and agentic work
- API Free tier: not supported
GPT-6 Astra specifications at a glance
Astra accepts text and image input and returns text. It provides a 1,050,000-token context window, up to 128,000 output tokens, and an April 30, 2026 knowledge cutoff. Audio and video input are not supported by the model itself.
The Responses API is the recommended interface for tool-based applications. Astra supports streaming, function calling, Structured Outputs, web search, file search, image generation, code interpreter, hosted shell, computer use, apply patch, and skills through supported API tools.
| Item | GPT-6 Astra |
|---|---|
| Model ID | gpt-6-astra |
| Context window | 1,050,000 tokens |
| Maximum output | 128,000 tokens |
| Knowledge cutoff | April 30, 2026 |
| Reasoning effort | low, medium, high, xhigh, max |
GPT-6 Astra API pricing
Standard API pricing is $10.00 per million input tokens, $1.00 per million cached input tokens, $12.50 per million cache-write tokens, and $50.00 per million output tokens. These rates exclude separate tool-call charges.
Prompts longer than 272K input tokens use long-context pricing for the full request: input and cache rates are multiplied by 2, while output is multiplied by 1.5. Batch and Flex are priced at 50% of Standard; Fast mode is 2 times the applicable Standard rate. Always verify the live invoice rules on OpenAI's pricing page.
| Token type | Standard price per 1M tokens |
|---|---|
| Input | $10.00 |
| Cached input | $1.00 |
| Cache writes | $12.50 |
| Output | $50.00 |
See Astra alongside GPT-5.6 Sol, Terra, and Luna.
Make your first GPT-6 Astra API request
Create a key on the OpenAI Platform, store it in the OPENAI_API_KEY environment variable, and use the Responses API. Never place a production key in browser code or a public repository.
Start with low or medium reasoning effort for routine work, then measure quality and cost before moving to high, xhigh, or max. Astra does not support the none reasoning setting.
from openai import OpenAI
client = OpenAI()
response = client.responses.create(
model="gpt-6-astra",
input="Review this deployment plan and identify the three highest risks.",
reasoning={"effort": "medium"},
)
print(response.output_text)Python example using the recommended Responses API.
import OpenAI from "openai";
const openai = new OpenAI();
const response = await openai.responses.create({
model: "gpt-6-astra",
input: "Review this deployment plan and identify the three highest risks.",
reasoning: { effort: "medium" },
});
console.log(response.output_text);TypeScript example. The SDK reads OPENAI_API_KEY from the environment.
GPT-6 Astra vs GPT-5.6 Sol, Terra, and Luna
Choose Astra when task completion quality matters more than per-token price: long agent trajectories, computer workflows, complex research, demanding coding, and professional documents. GPT-5.6 Sol remains a lower-priced flagship option, Terra balances capability and cost, and Luna is designed for high-volume cost-sensitive work.
Do not choose from benchmark headlines alone. Build a representative evaluation set and compare successful-task cost, latency, tool-call reliability, and human review time. A model with a higher token rate can still cost less per completed task if it needs fewer retries or fewer output tokens.
| Model | Input / output per 1M | Best starting point |
|---|---|---|
| GPT-6 Astra | $10 / $50 | Hard end-to-end and agentic work |
| GPT-5.6 Sol | $4 / $20 | Flagship quality at a lower token rate |
| GPT-5.6 Terra | $2 / $12 | Balanced production workloads |
| GPT-5.6 Luna | $0.20 / $1.20 | High-volume, cost-sensitive tasks |
New agent features: async tools and mid-turn steering
Async tool calling allows Astra to continue reasoning or handle independent work while a slow tool runs. Your application still executes the tool and must return the result using the original call ID. Use it only where work is genuinely independent; otherwise you can create race conditions or use stale assumptions.
With a WebSocket connection, mid-turn steering lets a user add a correction or new requirement without restarting the whole task. Configuration-update items can also change reasoning effort during a conversation while preserving cache-friendly prompt prefixes.
Migration checklist for GPT-6 Astra
Changing only the model string is not enough for a reliable migration. OpenAI recommends the Responses API for tool calling. Remove unsupported sampling fields such as temperature, top_p, and top_logprobs, and replace none or minimal reasoning with low.
Re-run safety, latency, cost, and tool-use evaluations before changing production traffic. If your prompts rely on implicit behavior, make task boundaries and stopping conditions explicit. Astra follows repository skills and instruction files closely, so audit files such as AGENTS.md before granting computer or shell access.
- Use gpt-6-astra and prefer the Responses API
- Replace none or minimal reasoning effort with low
- Remove unsupported sampling and log-probability parameters
- Test long-context cost above 272K input tokens
- Audit tool permissions, skills, and repository instructions
- Keep a GPT-5.6 fallback until production evaluations pass
Benchmarks and safety limits
OpenAI reports 59.3% on Agents' Last Exam, 72.6% on OSWorld 2.0, 92.7% on ScreenSpot-Pro, and 41.4% on AutomationBench. These figures describe controlled evaluations, not a guarantee for your application, and some comparisons use different harnesses or unavailable third-party scores.
OpenAI classifies Astra at the Critical cybersecurity capability level under its Preparedness Framework and applies stronger deployment safeguards. Use the model only for authorized work, keep human review around high-impact actions, apply least-privilege tool permissions, and log complete task trajectories.
FAQ
What is the GPT-6 Astra API model ID?
Use gpt-6-astra. Confirm that the model is available to your organization before routing production traffic.
How much does GPT-6 Astra cost?
Standard API rates are $10 per million input tokens, $1 cached input, $12.50 cache writes, and $50 output. Long-context, tools, Batch, Flex, and Fast pricing can change the final cost.
Is GPT-6 Astra available on the API Free tier?
No. The official model page lists the Free API tier as unsupported. ChatGPT plan availability and API billing are separate.
Does GPT-6 Astra support temperature?
OpenAI's migration guidance says to remove temperature, top_p, and top_logprobs. Control reasoning with reasoning.effort instead.
Should every application migrate from GPT-5.6 to Astra?
No. Astra targets the hardest end-to-end work and has higher token rates. Evaluate it against GPT-5.6 Sol, Terra, and Luna on your own quality, latency, and successful-task cost targets.
Related in This Series
Related Providers
Sources
- GPT-6 Astra: A new generation of intelligenceOpenAI · Checked 2026-09-09
- GPT-6 Astra model referenceOpenAI Developers · Checked 2026-09-09
- GPT-6 Astra model guidance and migrationOpenAI Developers · Checked 2026-09-09
- GPT-6 Astra safety overviewOpenAI · Checked 2026-09-09