OpenAI sells the GPT family as a hosted API. On today's feed the current general models are GPT-6.2 Astra, GPT-6.2 Sol, and the cheaper GPT-6 Luna, plus older GPT-5 snapshots, audio models, and image models. Astra Pro is listed as the same model as Astra with a higher reasoning mode, not as a separate brain.
Flagship · $8.00 in · $40.00 out · 1,050,000 tokens
Guide note · 19 Nov 2026
OpenRouter describes GPT-6.2 Astra as OpenAI's flagship for the hardest projects, including engineering and scientific research. Use it when a wrong answer is expensive to redo. The feed lists Astra Pro as the same model with a higher reasoning mode, and warns that pro mode spends more, even though the published per-token rates match.
Listed as the flagship for hard engineering and research
A flagship pass is for a design you will live with: a migration plan, a review of a security design, or the method section of a paper. Put the documents in the request. The feed does not publish a knowledge cutoff, so treat Astra as a reader of what you supply.
GPT-6.2 Astra is listed at $8.00 input and $40.00 output per million tokens, with a 1,050,000-token context window.
Long research report from papers you already have
File input and a million-token-class window can hold a memo plus the sources you upload. Ask for gaps and contradictions, then check every claim. This is a reading aid, not a library.
GPT-6.2 Astra is listed at $8.00 input and $40.00 output per million tokens, with a 1,050,000-token context window.
The tasks are editorial illustrations, not measured benchmarks. The prices in them are read from today's feed.
Everyday · $2.00 in · $10.00 out · 1,050,000 tokens
Guide note · 19 Nov 2026
OpenRouter positions GPT-6.2 Sol under Astra as the efficient model for everyday work. It is the reviewed default when you want a strong general model and do not need the flagship.
A form, a list, and a bit of saved state is everyday engineering. Give Sol a short spec and read the diff. It can also look at a screenshot of the page you want.
GPT-6.2 Sol is listed at $2.00 input and $10.00 output per million tokens, with a 1,050,000-token context window.
Rewrite one important customer email
A single delicate note — a refund that went wrong, a delay you have to explain — is worth a mid-tier pass. You can attach a screenshot of the thread.
GPT-6.2 Sol is listed at $2.00 input and $10.00 output per million tokens, with a 1,050,000-token context window.
The tasks are editorial illustrations, not measured benchmarks. The prices in them are read from today's feed.
OpenRouter describes GPT-6 Luna as the fast, cost-efficient GPT-6 model for high-volume work such as chat, classification, and light agent steps. It is the reviewed cheap OpenAI option.
Sorting mail into refund, bug, sales, or other, and writing a one-line routing note, is the job the listing describes. Keep the label set small and sample the output.
GPT-6 Luna is listed at $0.10 input and $0.50 output per million tokens, with a 1,050,000-token context window.
Classify support tickets as they arrive
A short ticket plus a fixed menu of labels is a latency-sensitive call you make all day. Luna lists optional reasoning, so you can turn the extra thinking down.
GPT-6 Luna is listed at $0.10 input and $0.50 output per million tokens, with a 1,050,000-token context window.
The tasks are editorial illustrations, not measured benchmarks. The prices in them are read from today's feed.
For most people, start with GPT-6.2 Astra, then drop to the next model on this page when the task is routine. Read the cheaper-model note under each example.
Are the examples benchmarks?
No. They are concrete tasks used to explain fit. The prices next to them are live figures from the feed. The judgment is a guide note.