SunsettingAI Dev Tools

GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 pro, o3, o3-pro (2025 snapshots)

by OpenAI

Announced

2026-06-11

Complete

2026-12-11

Refund window closes

—

Refund status

No refund applicable

On June 11, 2026 OpenAI notified developers that six older GPT-5 and o3 snapshots will be removed from the API on December 11, 2026: `gpt-5-2025-08-07`, `gpt-5-mini-2025-08-07`, `gpt-5-nano-2025-08-07`, `gpt-5-pro-2025-10-06`, `o3-2025-04-16` and `o3-pro-2025-06-10`. These are the default snapshots behind the `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `gpt-5-pro`, `o3` and `o3-pro` model pages. The deprecations page names a GPT-5.6 replacement for each: `gpt-5.6-sol` for gpt-5 and o3, `gpt-5.6-terra` for gpt-5-mini, `gpt-5.6-luna` for gpt-5-nano, and `gpt-5.6-sol` with `reasoning.mode: pro` for gpt-5-pro and o3-pro. Read from OpenAI's model pages on October 10, 2026, every one of these swaps raises the listed input and output price except the two pro rows. The steepest is gpt-5-mini to gpt-5.6-terra, where listed input goes up 8x and output 6x. For a pipeline that uses these models to write video prompts, shot lists or captions, the bill is the first thing to recheck.

Refund flow

1
Not applicable. These are pay-per-token API models with no pre-paid balance tied to a model string, so there is no refund category to open.
2
If you are billed for usage of any of these six snapshots dated after December 11, 2026, treat it as a billing error rather than a refund request: open a ticket at help.openai.com with the invoice line and the request IDs.
3
Re-budget before you switch. Per-1M-token prices listed on OpenAI's model pages on October 10, 2026 (input / cached input / output): gpt-5 $1.25 / $0.125 / $10, gpt-5-mini $0.25 / $0.025 / $2, gpt-5-nano $0.05 / $0.005 / $0.40, o3 $2 / $0.50 / $8, gpt-5-pro $15 / none / $120, o3-pro $20 / none / $80. Replacements: gpt-5.6-sol $4 / $0.40 / $20, gpt-5.6-terra $2 / $0.20 / $12, gpt-5.6-luna $0.20 / $0.02 / $1.20, each with cache writes billed at 1.25x the uncached input rate.
4
What that works out to: gpt-5 to gpt-5.6-sol is 3.2x input and 2x output. gpt-5-mini to gpt-5.6-terra is 8x input and 6x output. gpt-5-nano to gpt-5.6-luna is 4x input and 3x output. o3 to gpt-5.6-sol is 2x input and 2.5x output, with cached input about 20% cheaper. On the pro rows the per-token rate falls, because pro mode bills at gpt-5.6-sol's standard rates, but OpenAI says pro mode does more model work and uses more tokens, so the per-job cost has to be measured, not assumed.
5
One more date to note: the gpt-5.6-sol page calls $4 and $20 promotional pricing, available at least through November 21, 2026, which is before the December 11 shutdown. The page describes those prices as a 20% input and 33% output reduction, which implies undiscounted rates of $5 and $30 if the promotion ends.

Migration path

Recommended successor

gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna

Why this is the closest fit

They are the replacements OpenAI names on each June 11, 2026 deprecation row, and the gpt-5.6-sol page says it roughly corresponds to the unsuffixed tier of earlier GPT-5 families. All three list Chat Completions and Responses as supported endpoints, so the request shape you use for gpt-5, gpt-5-mini, gpt-5-nano or o3 is still available.

What differs from the original

Four things change besides the model string. First, reasoning effort: the gpt-5 page lists minimal, low, medium and high, while the GPT-5.6 pages list none, low, medium, high, xhigh and max with medium as the default, so minimal is not in the new list and any call that sends it needs a new value. Second, pro is now a mode, not a model: gpt-5-pro and o3-pro become gpt-5.6-sol with reasoning.mode set to pro, which OpenAI documents for the Responses API, matching the old pro models, which were Responses-only. Third, multi-turn context: OpenAI's reasoning guide says GPT-5.6 models default to rendering reasoning from earlier turns, where older models did not, and reasoning.context selects either behavior. Fourth, context size: gpt-5, mini and nano list a 400,000-token window with 272,000 max input, o3 lists 200,000, and the replacements list 1,050,000, but a prompt over 272K input tokens is billed at 2x input and 1.5x output for the whole request. Fine-tuning is listed as not supported on all six retiring models and all three replacements.

Alternatives by use case

You run high-volume prompt expansion, tagging or captioning on gpt-5-mini

Price the terra swap first, because it is the largest jump on the list. The gpt-5.6-luna page lists $0.20 input and $1.20 output, which is below gpt-5-mini's $0.25 and $2 on both, so test whether luna is good enough for the job before accepting terra's rate

You use gpt-5 or o3 to draft video scripts or shot lists

Move to gpt-5.6-sol and set reasoning effort explicitly, since the new default is medium. Budget at the undiscounted $5 and $30 rather than the promotional $4 and $20, since the promotion is only stated through November 21, 2026

Your code sends reasoning effort minimal for speed

Change it to none or low before switching, because minimal is not a listed value on the GPT-5.6 pages

You call gpt-5-pro or o3-pro for hard one-off reviews

Run a few real jobs on gpt-5.6-sol with reasoning.mode pro and compare total billed tokens, since the per-token rate drops but the token count per job is expected to rise

What this tool meant

gpt-5-2025-08-07 is the snapshot that launched the GPT-5 family, and o3 was the reasoning model the o3 page itself now describes as succeeded by GPT-5. Both leave the API about sixteen months and twenty months after their snapshot dates, on the six months' notice OpenAI promises for generally available models. The pattern repeats from the October 23, 2026 and April 1, 2027 batches: the named replacement covers capability, not price. Here the listed input and output prices rise on every non-pro row, so the cheapest migration may be a different tier than the one OpenAI maps you to.