GPT-6.2 Astra isn’t free. As of November 18, 2026, every route to it costs money: ChatGPT Work and Codex on Plus, Pro, Enterprise or Edu, GitHub Copilot Pro+ and above, or the API at $8 per 1M input tokens and $40 per 1M output tokens. The cheapest way to test it is a small, capped API run that costs a dollar or two.
OpenAI shipped GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. This guide checks each place you might expect a free tier, shows which ones carry the model, and gives you a test plan that keeps your first bill under $10.
Is GPT-6.2 Astra free?
GPT-6.2 Astra has no free tier. OpenAI’s launch post names Plus, Pro, Enterprise and Edu as the ChatGPT plans that get it, the API has no free usage for it, GitHub Copilot limits it to paid Pro+, Max, Business and Enterprise seats, and OpenRouter lists no :free variant. Every path below was checked on November 18, 2026.
| Route | GPT-6.2 Astra? | What you pay |
|---|---|---|
| ChatGPT Free and Go | Not in the rollout | n/a |
| ChatGPT Plus, Pro, Enterprise, Edu | Yes, in ChatGPT Work and Codex | Your plan fee |
| GitHub Copilot Free and Pro | Not listed | n/a |
| GitHub Copilot Pro+, Max, Business, Enterprise | Yes | AI Credits at provider list price |
| OpenAI API | Yes | $8 in, $40 out per 1M tokens |
| OpenRouter | Yes, openai/gpt-6.2-astra |
$8 in, $40 out per 1M tokens |
There’s no free way in today. What you control is how little you spend to find out whether Astra is worth it.

Where GPT-6.2 Astra is available
ChatGPT Work and Codex
OpenAI’s launch announcement says both models are “available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users.” Free and Go aren’t on that list. The same post adds that “this rollout does not add either model to regular Chat”, so even on Plus you won’t find Astra in the normal chat picker.
Availability also “depends on your plan, workspace permissions, and rollout progress.” A Business Premium user on the forum reported no access yet, and a moderator called that normal. If you pay for Plus and don’t see it, update the Codex CLI and wait for the rollout.
GitHub Copilot
GitHub’s November 17 changelog makes both models available to Copilot Pro+, Max, Business and Enterprise users, billed at provider list pricing under usage-based billing. Copilot Free and Copilot Pro aren’t included.
The Copilot plans page lists Pro+ at $39 per user per month with $70 in monthly AI Credits ($39 base plus a $31 flex allotment), where one credit equals $0.01. Astra calls draw those credits down at list price, so Copilot isn’t a discount. It’s a convenient wrapper if you already pay for Pro+.
The OpenAI API
The GPT-6.2 Astra model page lists gpt-6.2-astra on the Responses, Chat Completions and Batch endpoints, with a 1,050,000-token context window. OpenAI’s pricing page shows no free tier, free credits or data-sharing program for it. The only “free” lines on that page are the moderation model and 1 GB of file search storage.
OpenRouter
OpenRouter’s model page lists openai/gpt-6.2-astra at $8 input and $40 output per 1M tokens across OpenAI, Azure and Amazon Bedrock endpoints, pay-as-you-go. There’s no free variant. Some providers on that page charge more, such as the EU in-region Azure endpoint at $9.60 and $48.
The cheapest way to test GPT-6.2 Astra
If you only want to know whether Astra beats your current model, you don’t need a subscription. A pay-as-you-go API run with a few guardrails is the cheapest route. Follow these steps:
- Write 10 to 20 prompts from your real work. Pick the tasks where your current model struggles. Generic benchmarks won’t tell you much.
- Run them on GPT-6.2 Sol first. Sol costs $2 in and $10 out, a quarter of Astra on every line, with the same context window and tools. If Sol passes, you may never need Astra.
- Set reasoning effort to
high, notmax. Astra rejectsnone, so every call reasons, and reasoning tokens bill as output. - Cap the answer with
max_output_tokens. This puts a hard ceiling on each call’s cost. Set it too low and a long reasoning pass can stop before the visible answer, so start around 4,000. - Keep prompts under 272K input tokens. Above that line, input doubles to $16 and output rises to $60 per 1M for the whole request.
- Send the eval through Batch if it can wait. Batch is 50% off: $4 in and $20 out for Astra.
- Set a monthly budget and a usage alert in the OpenAI dashboard before you start, so a runaway loop can’t surprise you.
Effort matters more than you’d expect. Simon Willison’s launch-day test drew the same image at three effort levels. Astra cost 11.55 cents at high, 17.70 cents at xhigh and 88.71 cents at max. Sol cost 2.41, 6.74 and 13.69 cents. Going from high to max on Astra made one answer almost eight times pricier.
What a GPT-6.2 Astra test actually costs
Here’s what a few test sizes cost at list price, assuming 2,000 input tokens per prompt and 1,500 output tokens, reasoning included. The “long reasoning” row uses 8,000 output tokens per prompt, closer to what max effort produces.
| Test run | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
20 prompts, high effort |
$1.52 | $0.38 |
| 20 prompts, long reasoning | $6.72 | $1.68 |
| 100-prompt eval, Standard | $7.60 | $1.90 |
| 100-prompt eval, Batch | $3.80 | $0.95 |
The math is simple. Twenty prompts send 40,000 input tokens ($0.32) and get 30,000 output tokens back ($1.20). Output is five times the input rate, so the output side drives almost every row.

That’s the real answer to “is it free”: no, but a useful first test costs about as much as a coffee. Your number will differ, so log usage.output_tokens on every call and multiply by $40 per 1M.
Which route is cheapest for you?
The best route depends on what you already pay for.
| If you… | Cheapest way to try GPT-6.2 Astra |
|---|---|
| Already pay for ChatGPT Plus or Pro | Codex or ChatGPT Work, at no extra charge beyond your plan |
| Already pay for Copilot Pro+ or Max | Pick it in the Copilot model picker; it uses your monthly AI Credits |
| Have neither, and want to test via API | Pay-as-you-go OpenAI API with a small budget cap |
| Route traffic across several providers | OpenRouter, at the same $8 and $40 list price |
| Only use free plans | Not possible today; test GPT-6.2 Sol on the API instead |
If you’re on a free plan and want a free model to compare against, our guide to whether GPT-6 Sol is free covers the earlier Sol release, and GPT-6 Sol pricing explains why Sol’s list price hasn’t moved across three generations.
Should you test GPT-6.2 Sol first?
For most teams, yes. OpenAI gives GPT-6.2 Sol the same 1,050,000-token context, the same 128,000-token output cap and the same tool support as Astra, at one quarter of the price.
Early reviews of Astra are mixed. Willison notes that Astra “isn’t scoring quite as high as Fable 5.2 on various benchmarks”, and he found that neither GPT-6.2 model produced a good enough result on his mech animation prompt. GitHub, on the other hand, says Astra caught more issues than GPT-6 Astra in code review while using fewer tokens.
That split is exactly why a cheap, task-specific test beats reading launch posts. Run your 20 prompts on Sol. Rerun only the failures on Astra. If Astra fixes them, you’ve found the tasks worth paying 4x for. If you’re coming from the previous generation, our GPT-6 Astra API guide shows the request shape, which carries over with a new model ID.
Let your coding agent run a capped GPT-6.2 Astra test with Apidog CLI
Running 20 prompts by hand and copying token counts into a spreadsheet is slow. Hand it to your coding agent instead, and let Apidog CLI run the requests, check the usage fields and fail the run if a call costs too much.
npm install -g apidog-cli@latest
apidog login --with-token <YOUR_TOKEN>
npx -y skills add https://apidog.com # pick your agent when prompted
Then give your agent one prompt:
In Apidog project <id>, create a test case for POST /v1/responses. Send the
same prompt to model gpt-6.2-sol and then gpt-6.2-astra, both with reasoning
effort "high" and max_output_tokens 4000. Assert each call returns 200,
usage.input_tokens stays under 272000 and usage.output_tokens stays under
4000. Run it against the openai environment and report output tokens and
estimated cost per call at $10 and $40 per 1M output tokens.
The agent looks up the payload shape, validates it, creates the case and runs it:
apidog cli-schema get test-case-create
apidog cli-schema validate test-case-create --file case.json
apidog test-case create --project <id> --file case.json
apidog run --test-case <caseId> --project <id> -e <envId> -r cli,json
The JSON report gives you token counts per call, so the cost comparison between Sol and Astra comes from your own prompts, not a benchmark chart. Add --upload-report if you want to review the run in the Apidog app afterward. Apidog CLI in Codex covers the agent setup if you’re testing from Codex, which is also where Plus users get Astra. Apidog’s Free plan covers up to 4 users.
FAQ
Is GPT-6.2 Astra free in ChatGPT? No. OpenAI’s announcement makes it available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. Free and Go aren’t listed, and the rollout doesn’t add it to regular Chat.
Is there a free GPT-6.2 Astra API tier? No. The API charges $8 per 1M input tokens and $40 per 1M output tokens on Standard, and OpenAI’s pricing page lists no free tier or free credits for the model.
Can I use GPT-6.2 Astra with GitHub Copilot Free? No. GitHub’s changelog limits it to Copilot Pro+, Max, Business and Enterprise, billed at provider list pricing through AI Credits.
Does OpenRouter have a free GPT-6.2 Astra model? No. OpenRouter lists openai/gpt-6.2-astra at $8 input and $40 output per 1M tokens, with no :free variant as of November 18, 2026.
What’s the cheapest way to test GPT-6.2 Astra? Run 10 to 20 real prompts through the API at high effort with max_output_tokens set. That costs about $1.52 on Astra, or $0.38 if you start with GPT-6.2 Sol. Batch halves both.
Is GPT-6.2 Sol free? No. Sol costs $2 input and $10 output per 1M tokens on the API and reaches ChatGPT and Copilot on the same paid plans as Astra.
