Tally
Numbers / GPT nano (latest) · API input price / Recipe

Recipe: GPT nano (latest) · API input price

What is the OpenAI API price per million input tokens of OpenAI's latest nano model? Track the newest model in that line over time, so the history covers the price of each earlier model in the line at its launch. Today: $0.2 /MTok. This is version 1.

Source

https://developers.openai.com/api/docs/pricing.md

OpenAI API pricing page, markdown version: https://developers.openai.com/api/docs/pricing.md (Standard tier table). Cross-check against the model page https://developers.openai.com/api/docs/models/gpt-5.4-nano.md and the model catalog https://developers.openai.com/api/docs/models.md, which shows which nano models exist. Fallback search: "OpenAI introducing nano model price per 1M input tokens".

Read

In the catalog, find the newest model whose ID ends in "-nano". Today that is gpt-5.4-nano (snapshot gpt-5.4-nano-2026-03-17). On the pricing page, in the first Standard pricing table, read the Input column for that row: $0.20 per 1M tokens. Ignore Batch and Flex ($0.10), cached input ($0.02), output ($1.25) and the 10% regional-processing uplift.

History

I took each launch price from OpenAI's own launch posts. GPT-4.1 nano was the first nano model and launched on 2025-04-14 at $0.10 (openai.com/index/gpt-4-1/). GPT-5 nano launched on 2025-08-07 at $0.05 (openai.com/index/introducing-gpt-5-for-developers/). GPT-5.4 nano launched on 2026-03-17 at $0.20 (openai.com/index/introducing-gpt-5-4-mini-and-nano/). No other model is named nano: there is no 5.1, 5.2, 5.5 or 5.6 nano. The current pricing page still lists all three models at their launch prices, so none of them changed price after launch. To extend this, add a point only if OpenAI launches a new model with "nano" in its ID, or changes gpt-5.4-nano's price.

Update

This changes only with an event: a new nano launch or a price change, both of which are rare. Re-read the pricing page and the model catalog once a month. The nano name looks retired: gpt-5.4-nano is marked Deprecated, and its listed replacement is gpt-6-luna, with shutdown on Apr 1, 2027. After that the row may disappear from the pricing page, and the last value should be kept.

Checks

Read the Standard input price, not Batch/Flex ($0.10) or cached input ($0.02).

Make sure the row is the newest -nano model, not gpt-5-nano ($0.05) or gpt-4.1-nano ($0.10).

Do not swap in a Luna model (gpt-6-luna $0.10, gpt-5.6-luna $0.20) unless the tracked line is redefined as "nano successors". OpenAI lists Luna as the replacement in its deprecations table, but those models are not named nano.

Prices are per 1M tokens and in USD. Ignore the 10% data-residency uplift.

How this version was produced: 5 steps by an agent, 6 checks
  1. Read developers.openai.com/api/docs/pricing
  2. Read developers.openai.com/api/docs/pricing.md
  3. Read developers.openai.com/api/docs/models.md
  4. Read developers.openai.com/api/docs/models/gpt-5.4-nano.md
  5. Searched the web: “Introducing GPT-5.4 mini and nano pricing per million input tokens”
  6. ✓ The recipe produced a value
  7. ✓ It still answers this number's question
  8. ✓ The evidence page was actually read in this run
  9. ✓ The value can be read in the source text
  10. ✓ No unexplained jump from the current value
  11. ✓ Past points can be read in the pages they cite · 3 of 3 kept.

Agent's note: The newest model OpenAI has named "nano" is gpt-5.4-nano, launched March 17, 2026 at $0.20 per 1M input tokens. Before it came gpt-5-nano at $0.05 (Aug 7, 2025) and gpt-4.1-nano at $0.10 (Apr 14, 2025). Each price comes from OpenAI's launch post, and the current pricing page lists all three unchanged. One catch: OpenAI seems to have dropped the nano name. gpt-5.4-nano is now marked Deprecated, and the deprecations page lists gpt-6-luna ($0.10 per 1M input) as its replacement. If you count the Luna models as the nano line, the current value would be $0.10 instead.

Model: claude-opus-5-5

Improve this recipe

Edit the text and press the button. An agent follows your version once; if the value it gets can be read in the source and is not an unexplained jump, your version goes live straight away. Otherwise it is kept in the history as not adopted, with the reason.

Versions

Back to the number