Budgeting the migration
OpenAI's named replacement for GPT-3.5 Turbo is GPT-5.4 mini, priced at $0.75 input / $4.5 output per MTok (verified 2026-09-07). What common volumes cost at those rates, assuming a 75% input / 25% output token split, before caching or batch discounts:
| Volume | Split | Cost at GPT-5.4 mini rates |
|---|---|---|
| 100K tokens in a day | 75,000 in + 25,000 out | $0.1688 |
| That pace over a 30-day month (3M tokens) | 2.25M in + 750K out | $5.06 |
| 1M tokens in a month | 750,000 in + 250,000 out | $1.69 |
| Heavy month — 12M tokens | 9M in + 3M out | $20.25 |
Token counts are not portable across model families — tokenizers differ — so treat these as rate-card numbers and re-measure with your own payloads after migrating.
Frequently asked questions
When does GPT-3.5 Turbo shut down?
OpenAI has scheduled GPT-3.5 Turbo for shutdown on 2026-10-23. Shutdown 2026-10-23 → migrate to GPT-5.4 mini.
What replaces GPT-3.5 Turbo?
The named migration target is GPT-5.4 mini, currently priced at $0.75 per 1M input tokens and $4.5 per 1M output tokens (verified 2026-09-07).
What will GPT-5.4 mini cost to run?
At GPT-5.4 mini rates, 100K tokens a day (75% input / 25% output) costs about $0.1688 — roughly $5.06 over a 30-day month — and 1M input + 1M output tokens together cost $5.25. Straight rate-card math; token counts differ between model families, so re-measure with real payloads.
Why is no price listed for GPT-3.5 Turbo?
We publish only prices verified against the provider's official pricing page, and OpenAI's pricing page no longer lists current rates for GPT-3.5 Turbo — so this page tracks its status, shutdown date and migration path instead.