GPT-5.4 mini and nano are OpenAI's older small models, both with 400K-token context and no separate cache-write charge. Mini costs roughly 3.6–3.75× more per token; the bigger question for new projects is whether to use either.
GPT-5.4 nano is about 3.7× cheaper than mini on input and output. But GPT-6 Luna ($0.10/$0.50 per 1M) is cheaper than both and has a 1M context window, so stay on 5.4 only for prompts you have already validated there.
| GPT-5.4 mini | GPT-5.4 nano | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Tier | Budget | Budget |
| Input / 1M tokens | $0.75 | $0.20 |
| Output / 1M tokens | $4.50 | $1.25 |
| Cache hit / 1M | $0.075 | $0.02 |
| Cache write / 1M | — | — |
| Context window | 400K | 400K |
| Notes | Cheaper, well-defined tasks | Fastest, cheapest |
GPT-5.4 nano is 3.8× cheaper on input tokens ($0.20 vs $0.75 per 1M). GPT-5.4 nano is 3.6× cheaper on output tokens ($1.25 vs $4.50 per 1M).
| Workload | GPT-5.4 mini | GPT-5.4 nano | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $33.00 | $9.00 | GPT-5.4 nano saves $24.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $26.10 | $7.00 | GPT-5.4 nano saves $19.10 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $1.25 | $0.34 | GPT-5.4 nano saves $0.91 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
nano fails your quality bar on a well-defined task — structured extraction, tool calls, short reasoning — and you have production prompts tuned on GPT-5.4 that you are not ready to re-test.
you need the cheapest GPT-5.4-family option for high-volume classification, routing or tagging and are pinned to the 5.4 generation.
Mini only earns its premium if it removes work elsewhere: fewer retries, fewer escalations to a larger model, or fewer human reviews. Because mini costs about 3.7× nano per token, nano stays cheaper even if every one of its failures is retried on mini, unless failures are very frequent or expensive to catch downstream. Measure failure rates on your own data before paying the $4.50 output rate.
Cache hits are $0.075 on mini and $0.02 on nano per 1M, a 90% discount on both, and OpenAI lists no cache-write fee or long-context tier for either. The 400K window is the hard ceiling — there is no way to pay more for longer prompts.
GPT-6 Luna costs $0.10 input and $0.50 output, and GPT-5.6 Luna $0.20/$1.20; both have 1M-token context. For a new build there is little price reason to start on the 5.4 generation, and OpenAI's half-price Batch API applies across all of them.
Nano is $0.20 input and $1.25 output per 1M tokens versus $0.75 and $4.50 for mini — roughly 3.6–3.75× cheaper.
Yes. GPT-6 Luna is listed at $0.10 input and $0.50 output per 1M, below nano on both.
Both list 400K tokens.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.