GPT-5.4 Mini vs GPT-5.4 Nano
Green marks the cheaper side per row — informational only, never a verdict for your workload
| API status | Available | Deprecated — shuts down 2027-04-01 |
|---|---|---|
| Context window | 400,000 tokens | 400,000 tokens |
| Knowledge cutoff | Aug 31, 2025 | Aug 31, 2025 |
| Standard input | $0.75/MTok | $0.20/MTok |
| Standard cached input | $0.075/MTok | $0.02/MTok |
| Standard output | $4.50/MTok | $1.25/MTok |
| Cache writes | No surcharge | No surcharge |
| Reasoning effort levels | none (default), low, medium, high, xhigh | none (default), low, medium, high, xhigh |
| Faster tiers offered | Fast | None published |
| Long-context repricing rule | Not published | Not published |
| Build tier rate limit | 5,000 RPM · 2,000,000 TPM | 5,000 RPM · 2,000,000 TPM |
| Codex with ChatGPT sign-in | Retired 2026-08-31 | No retirement announced |
The budget end of the GPT-5.4 generation, and two models that are closer than their price gap suggests: the same context window, the same knowledge cutoff, the same effort levels with none as the default, and — since OpenAI's usage tiers changed on October 6, 2026 — identical rate-limit tables.
Price
Nano is cheaper than Mini on every published cell — input, cached input and output — at every tier both offer. Neither adds a cache-write surcharge.
Rate limits no longer differ
Both use the Luna-group table. Nano used to have a smaller token ceiling and batch queue at Tier 1; that row went with Tier 1 itself when OpenAI moved to Build, Launch and Grow, and above it the two tables were already the same.
Fast mode
Mini publishes a Fast tier. Nano doesn't — Standard is its fastest paid option.
Availability changed for one of them
GPT-5.4 Mini left Codex for ChatGPT sign-in on August 31, 2026, with GPT-6 Luna as OpenAI's named replacement. Nano wasn't on that list — but on the API it is now the one on its way out: OpenAI deprecated it on October 1, 2026 and shuts it down on April 1, 2027, again naming GPT-6 Luna as the replacement. Mini has no announced API shutdown.
How to decide
Between these two: Nano only for work you will have moved before its API shutdown; Mini if you need Fast mode at all, or a model with no shutdown date. But the more useful comparison is now against GPT-6 Luna, which undercuts both on every Standard price cell with a larger window and a newer cutoff.
Verified 2026-10-28 against CodexHow facts module (src/data/facts/) — see /about/#accuracy.