AI Model Watch
Models / OpenAI

GPT-3.5 Turbo

Retired

gpt-3.5-turbo-0125, including the gpt-3.5-turbo alias and gpt-3.5-turbo-completions, was shut down on 2026-10-23 and OpenAI names gpt-5.6-terra as the replacement. Until 2026-10-24 the row was held at deprecated on OpenAI's own deprecations page, section '2026-04-22: Legacy GPT model snapshots', which stated the shutdown date and the recommended replacement as columns of a table; the date passing on 2026-10-23 was not treated as evidence on its own. Rates read the same day from developers.openai.com/api/docs/pricing (Standard tier), which priced deprecated models; on the 2026-10-24 read the pricing page's own data payload still carries these same rates, so they are kept as the last published ones. Standard-tier rates: input $0.50, output $1.50 per 1M, with no cached-input rate published; OpenAI prices the alias and the dated snapshot identically. Specs are from developers.openai.com/api/docs/models/gpt-3.5-turbo: 16,385 context window, 4,096 max output tokens, knowledge cutoff Sep 01, 2021, text only in and out. released is null on purpose: the snapshot id encodes 01-25 with no year and OpenAI publishes no release date for it, so dating it would mean supplying the year ourselves. The separately priced gpt-3.5-turbo-1106 and gpt-3.5-turbo-instruct rows appear in no deprecation table and are not carried yet. A fine-tuned variant, ft-gpt-3.5-turbo, carries the same shutdown date and is not covered by this row. THE DELAY IS OURS, NOT A PROVIDER EVENT: the announcement, the date and the replacement are all OpenAI's, published 2026-04-22; this catalog simply never carried the row. It was added 2026-09-21, 32 days before the shutdown. CONFIRMED RETIRED 2026-10-24. Signal (first-party, past tense): the same section of developers.openai.com/api/docs/deprecations now states "Access to these models was shut down on October 23, 2026." and lists gpt-3.5-turbo-0125 | gpt-3.5-turbo, gpt-3.5-turbo-completions in its table with substitute gpt-5.6-terra - an executed shutdown, not a scheduled one. The substitute is unchanged.

GPT-3.5 Turbo by OpenAI costs $0.5 per 1M input tokens and $1.50 per 1M output tokens ($0.75/1M blended), with a 16K (16.385-token) context window. It has been retired (shut down 23 Oct 2026); the recommended migration is GPT-5.6 Terra.

Last verified: 24 Oct 2026 · sourced from official provider documentation

Retired on 23 Oct 2026 · deprecated since 22 Apr 2026

Lifecycle warning

This model has been retired. Recommended migration: GPT-5.6 Terra.

Read the GPT-3.5 Turbo migration guide → replacement & cheapest alternatives

Provider
OpenAI
Status
Retired
Input price
$0.5 / 1M tokens
Output price
$1.50 / 1M tokens
Cached input
—
Blended price
$0.75 / 1M tokens
Context window
16.385 tokens (16K)
Max output
4096 tokens
Modality
text
Open weights
No — API only
Knowledge cutoff
2021-09-01
Released
—
API string
gpt-3.5-turbo-0125

Source: OpenAI official documentation ↗

See also

Explore more

All OpenAI models→

Prices, context windows and lifecycle status for every OpenAI model we track.

Compare models head-to-head→

Side-by-side price, context and status for the models developers weigh against each other.

FAQ

GPT-3.5 Turbo — questions & answers

How much does GPT-3.5 Turbo cost?

GPT-3.5 Turbo is priced at $0.5 per 1M input tokens and $1.50 per 1M output tokens ($0.75/1M blended at a 3:1 input-to-output ratio).

What is the context window of GPT-3.5 Turbo?

GPT-3.5 Turbo has a 16.385-token context window (16K), with up to 4096 output tokens per request.

Is GPT-3.5 Turbo still available?

GPT-3.5 Turbo has been retired and is no longer available. Its scheduled shutdown date is 23 Oct 2026. The recommended migration is GPT-5.6 Terra.

Does GPT-3.5 Turbo support image or vision input?

Our catalog lists GPT-3.5 Turbo's modalities as text; image/vision input is not among them.

Can I self-host GPT-3.5 Turbo?

No — GPT-3.5 Turbo is proprietary. Its weights are not publicly released, so it is available only through OpenAI's API (or hosted partners), not for self-hosting.

What is the API model string for GPT-3.5 Turbo?

The API model identifier for GPT-3.5 Turbo is "gpt-3.5-turbo-0125" when calling OpenAI's API.

What is GPT-3.5 Turbo's knowledge cutoff?

GPT-3.5 Turbo's training knowledge cutoff is 1 Sep 2021.