OpenAI · OpenAI’s own facts

6.2 Astra vs 6 Astra (2026): GPT-6.2 Astra against GPT-6 Astra, price, specs and OpenAI’s benchmarks

GPT-6.2 Astra, released on November 17, 2026, costs $8 per million input and $40 per million output tokens against $10 and $50 for GPT-6 Astra, has the same 1,050,000-token context window and a later knowledge cutoff (May 18, 2026 against April 30, 2026), and leads GPT-6 Astra on most of the benchmarks in OpenAI’s own system card.

Last updated:

Use it now

Chat

The chat starts with GPT-6.2 Astra; switch to GPT-6 Astra in the model list to ask it the same thing.

≈ 11 credits per reply · See all models

New accounts get 5 free credits once, valid for 7 days. They work on the lower-priced models here; the priciest ones need credits you bought.

Start chatting — 5 free credits

Per million tokens

Price per million tokens

checked 2026-11-19
-InputOutputOpenRouter id
GPT-6.2 Astra$8$40openai/gpt-6.2-astra
GPT-6 Astra$10$50openai/gpt-6-astra

OpenAI’s own facts

Side by side

-GPT-6.2 AstraGPT-6 Astra
Released
November 17, 2026 (OpenAI’s API changelog and community post)Source: developers.openai.com · checked 2026-11-19
Listed on OpenRouter since September 4, 2026; OpenAI’s model page shows no release dateSource: openrouter.ai · checked 2026-11-19
How OpenAI describes it
"Our most capable model for your hardest projects, from start to finish"Source: developers.openai.com · checked 2026-11-19
"Our flagship model for complex reasoning and coding" (OpenAI’s models page)Source: developers.openai.com · checked 2026-11-19
Price per 1M tokens (input / output)
$8 / $40 (cached input $0.40)Source: developers.openai.com · checked 2026-11-19
$10 / $50 (cached input $1)Source: developers.openai.com · checked 2026-11-19
Prompts over 272K input tokens
Input and cache rates 2x and output 1.5x for the full request ($16 / $60)Source: developers.openai.com · checked 2026-11-19
Input and cache rates 2x and output 1.5x for the full request ($20 / $75, our arithmetic)Source: developers.openai.com · checked 2026-11-19
Batch, Fast and Ultrafast
Batch $4 / $20; Fast $16 / $80; Ultrafast $24 / $120 until February 28, 2027, then $48 / $240Source: developers.openai.com · checked 2026-11-19
Batch and Flex 50% of Standard; Fast mode 2x; Ultrafast not listed on the pageSource: developers.openai.com · checked 2026-11-19
One request, 10,000 tokens in and 1,000 out (our arithmetic)
About $0.12Source: developers.openai.com · checked 2026-11-19
About $0.15Source: developers.openai.com · checked 2026-11-19
Context / max output
1,050,000 / 128,000 tokens (up to 922,000 input)Source: developers.openai.com · checked 2026-11-19
1,050,000 / 128,000 tokens (up to 922,000 input)Source: developers.openai.com · checked 2026-11-19
Knowledge cutoff
May 18, 2026Source: developers.openai.com · checked 2026-11-19
April 30, 2026Source: developers.openai.com · checked 2026-11-19
Reasoning effort
The API rejects effort "none"; OpenAI’s guide suggests starting at mediumSource: developers.openai.com · checked 2026-11-19
low, medium, high, xhigh and maxSource: developers.openai.com · checked 2026-11-19
Endpoints
Chat Completions, Responses and BatchSource: developers.openai.com · checked 2026-11-19
Chat Completions, Responses and BatchSource: developers.openai.com · checked 2026-11-19
GPQA Diamond, no tools (accuracy)
97.0%Source: deploymentsafety.openai.com · checked 2026-11-19
96.0%Source: deploymentsafety.openai.com · checked 2026-11-19
MLE-Bench Revised (mean reference percentile)
96.06Source: deploymentsafety.openai.com · checked 2026-11-19
93.80Source: deploymentsafety.openai.com · checked 2026-11-19
NanoGPT (share of baseline training time saved)
42.87%Source: deploymentsafety.openai.com · checked 2026-11-19
7.73%Source: deploymentsafety.openai.com · checked 2026-11-19
PostTrainBench Lite (share of performance gap closed)
74.79%Source: deploymentsafety.openai.com · checked 2026-11-19
76.97%Source: deploymentsafety.openai.com · checked 2026-11-19
ExploitBench Internal Port, June–August 2026 (code execution rate)
53.5%Source: deploymentsafety.openai.com · checked 2026-11-19
31.5%Source: deploymentsafety.openai.com · checked 2026-11-19
Via OpenRouter
openai/gpt-6.2-astra at $8 / $40, 1,050,000-token contextSource: openrouter.ai · checked 2026-11-19
openai/gpt-6-astra at $10 / $50, 1,050,000-token contextSource: openrouter.ai · checked 2026-11-19
API id (maker)
gpt-6.2-astraSource: developers.openai.com · checked 2026-11-19
gpt-6-astraSource: developers.openai.com · checked 2026-11-19

Which one does the maker recommend?

  • In OpenAI’s system card (OpenAI’s numbers, not tested by us) GPT-6.2 Astra is ahead of GPT-6 Astra on seven of the nine tests shown there, by a wide margin on NanoGPT (42.87% against 7.73%) and ExploitBench Internal Port (53.5% against 31.5%), but GPT-6 Astra scores higher on PostTrainBench Lite (76.97% against 74.79%).Source: deploymentsafety.openai.com · checked 2026-11-19
  • OpenAI’s models page tells readers who are unsure to start with GPT-6 Astra, its flagship model for complex reasoning and coding; the GPT-6.2 Astra page calls the newer model its most capable for the hardest projects.Source: developers.openai.com · checked 2026-11-19

Sample answers (AI-generated)

Same question, both models

Sample answers are pending: they have not been generated yet.

FAQ

Questions

Is GPT-6.2 Astra better than GPT-6 Astra?

On OpenAI’s own system-card numbers, mostly yes: it scores higher on seven of the nine tests shown, for example 96.06 against 93.80 on MLE-Bench Revised and 53.5% against 31.5% on ExploitBench Internal Port, but GPT-6 Astra is higher on PostTrainBench Lite (76.97% against 74.79%). These are OpenAI’s numbers and we did not run them.

How much cheaper is GPT-6.2 Astra than GPT-6 Astra?

OpenAI lists $8 per million input tokens and $40 per million output tokens for GPT-6.2 Astra, against $10 and $50 for GPT-6 Astra, and cached input at $0.40 against $1. For a request with 10,000 tokens in and 1,000 out that is about $0.12 against $0.15 (our arithmetic). OpenRouter lists the same prices.

What are "openai 6.2 astra" and "gpt-astra 6.2"?

Both are searches for GPT-6.2 Astra, whose API id is gpt-6.2-astra; "ChatGPT 6.2 Astra" is the same model, which OpenAI’s community post says is in ChatGPT’s Work and Codex modes. The Sol sibling is GPT-6.2 Sol, at $2 / $10.

What changed in knowledge cutoff, context and reasoning effort?

The knowledge cutoff moves from April 30, 2026 (GPT-6 Astra) to May 18, 2026 (GPT-6.2 Astra). Both list a 1,050,000-token context, 922,000 tokens of input and 128,000 of output. GPT-6 Astra’s page lists reasoning effort low, medium, high, xhigh and max; for GPT-6.2 OpenAI says the API rejects "none" and suggests starting at medium.

Is GPT-6.2 Astra worth the price over GPT-6.2 Sol?

Astra costs four times Sol ($8 / $40 against $2 / $10). OpenAI’s system card shows Astra ahead on most of its tests, by a wide margin on NanoGPT, while Sol is ahead on SEC-Bench Pro. The cheaper path to try first is GPT-6.1 Sol vs GPT-6.2 Sol, and the full benchmark table is on GPT-6.2.

Can I chat with both on this site?

Yes. On this site a typical reply to GPT-6.2 Astra costs about 11 credits and a typical reply to GPT-6 Astra about 13 credits (a credit is about $0.01 in the $5 pack), billed by its real token cost, at least 1 credit. A cheaper OpenAI model is GPT-6.1 Terra.

Is this an official OpenAI page?

No. This page is unofficial and not affiliated with or endorsed by OpenAI. Every fact links to the OpenAI page we read.