Plan limits

What are the GPT-6 Sol and Luna rate limits on the OpenAI API?

On the OpenAI API, GPT-6 Sol runs from 500 requests and 500,000 tokens a minute at Tier 1 to 15,000 requests and 40,000,000 at Tier 5. Luna tops out at 30,000 and 180,000,000. Neither is available to the free tier.

Checked against the lab's own pages on September 22, 2026. Limits change often, so the date matters.

GPT-6 Sol and GPT-6 Luna went up on September 22, 2026, each with its own rate limit table on its model page. Both start the same way: the free tier can't call either one, and Tier 1 opens at 500 requests a minute and 500,000 tokens a minute on both.

Free

Not supported, on both models

Tier 1

500 RPM and 500,000 TPM on both, batch queue 1,500,000 on Sol and 5,000,000 on Luna

Tier 3

Sol 5,000 RPM and 2,000,000 TPM, Luna 5,000 RPM and 4,000,000 TPM

Tier 4

Sol 10,000 RPM and 4,000,000 TPM, Luna 10,000 RPM and 10,000,000 TPM

Tier 5

Sol 15,000 RPM and 40,000,000 TPM, Luna 30,000 RPM and 180,000,000 TPM

Sol's table is Astra's table

Row for row, GPT-6 Sol gets what the flagship gets. Every tier carries the same requests a minute, the same tokens a minute and the same batch queue limit as GPT-6 Astra, right down to 15,000,000,000 queued tokens at Tier 5. Sol costs $2 and $10 per million tokens against Astra's $10 and $50, so the cheaper model has the same ceiling on how fast you can push work through it.

How you climb the tiers, what the response headers tell you and what a 429 means don't change between models, and they're on our page for Astra's limits.

Luna is the table that's different

OpenAI calls Luna its most efficient model for focused, high-volume tasks, and the limits follow. Tokens a minute run ahead of Sol's from Tier 2 up: 2,000,000 against 1,000,000 at Tier 2, then 4,000,000 against 2,000,000 at Tier 3, then 10,000,000 against 4,000,000 at Tier 4. At Tier 5 it's 180,000,000 against 40,000,000, four and a half times as much, and requests a minute double to 30,000.

Requests a minute are identical on the two models everywhere below that: 500, then 5,000 for two tiers, then 10,000. Only the top row separates them.

The batch queue column moves in odd steps

The third column counts input tokens sitting in pending batch jobs, and it's the one people skip. At Tier 1, Luna allows 5,000,000 to Sol's 1,500,000. At Tier 3 it flips the other way, with Sol at 100,000,000 against Luna's 40,000,000. Tier 4 flips it back (1,000,000,000 on Luna, 200,000,000 on Sol), and at Tier 5 both sit at 15,000,000,000.

If you're sizing a batch job around a tier you haven't reached yet, that's the row to read twice, because it doesn't climb evenly on either model.

These are the tables as of September 22, 2026, the day both models were published. Both models carry the same 1,050,000-token context window and the same reasoning effort settings, so the choice between them is throughput and price.