GPT-3.5 Is Being Shut Down: Final Dates and What to Use Instead

OpenAI is retiring the gpt-3.5-turbo API on October 23, 2026. Here are the exact shutdown dates, what replaces GPT-3.5, and how to migrate before it's gone.

Quick answer. GPT-3.5 is being retired. OpenAI removed it from ChatGPT back in 2024, and the gpt-3.5-turbo API is scheduled to shut down on October 23, 2026 (gpt-3.5-turbo-instruct goes on September 28, 2026). If you’re still calling GPT-3.5, migrate to a current small model such as GPT-4o mini or the GPT-5 mini / nano tier — they’re cheaper and far more capable.

GPT-3.5 is the model that started the consumer-AI boom. It powered the original ChatGPT in late 2022 and, as gpt-3.5-turbo, became the cheap, fast workhorse that an entire generation of AI apps was built on. In 2026 it’s reaching the end of the road. Here are the exact shutdown dates, what GPT-3.5 actually was, how far it has fallen behind, and what to migrate to before it goes dark.

Is GPT-3.5 being discontinued?

Yes. GPT-3.5 has already been gone from the ChatGPT app since July 2024, when GPT-4o mini replaced it. The API is on a published shutdown clock. According to OpenAI’s deprecations schedule, the remaining GPT-3.5 models stop responding to API calls in autumn 2026 — which, with a few months’ notice, is exactly the window to plan a migration.

GPT-3.5 shutdown dates

ModelStatusFinal shutdown
gpt-3.5-turbo (→ -0125)DeprecatedOctober 23, 2026
gpt-3.5-turbo-1106DeprecatedOctober 23, 2026
Fine-tuned gpt-3.5-turbo modelsDeprecatedOctober 23, 2026
gpt-3.5-turbo-instructDeprecatedSeptember 28, 2026
-0301 / -0613 / -16k-0613RetiredSeptember 13, 2024

GPT-3.5 isn’t the only model retiring on October 23, 2026 — a batch of older models, including GPT-4 Turbo, sunsets the same day. Always confirm the current dates on OpenAI’s official deprecations page before you finalize a migration plan.

What was GPT-3.5?

GPT-3.5 was OpenAI’s mid-generation family that bridged GPT-3 and GPT-4. The key facts:

  • It launched ChatGPT on November 30, 2022 — the “free research preview” that put generative AI in front of the public.
  • gpt-3.5-turbo arrived as an API model on March 1, 2023 at $0.002 per 1K tokens — a roughly 10× price cut that made building AI products cheap enough to go mainstream.
  • Variants included gpt-3.5-turbo, the extended gpt-3.5-turbo-16k, and gpt-3.5-turbo-instruct for the older completions API.
  • Context window started at 4,096 tokens and reached 16,385 in the final -0125 snapshot. It was text-only, with a training cutoff around September 2021.

It was never the smartest model — it was the practical one. Good enough at chat, summarization, and simple tasks, and cheap enough to run at scale.

GPT-3.5 vs GPT-4 vs GPT-5: how far it has fallen behind

Every newer tier beats GPT-3.5 on capability, context, and — crucially — cost. The model that replaced it in ChatGPT, GPT-4o mini, already undercut GPT-3.5 on price while adding vision and a far larger context window.

ModelContextMultimodalStatus (mid-2026)
GPT-3.5 Turbo16KNo (text only)API shutdown Oct 23, 2026
GPT-4o mini128KYesReplaced 3.5 in ChatGPT (2024)
GPT-5 / 5 mini / nanoLargeYesCurrent default tier
GPT-5.5LargeYesNewest flagship (April 2026)

Want the full picture? Read our continuously-updated GPT-5.5 complete guide — capabilities, pricing, and how the current OpenAI lineup compares

What should you use instead of GPT-3.5?

OpenAI’s deprecation page doesn’t always spell out a one-to-one replacement, but for GPT-3.5’s “cheap and fast” role the practical successors are clear:

  • GPT-4o mini — the direct heir; multimodal, 128K context, and already cheaper than GPT-3.5 was.
  • GPT-5 mini / nano — the current small-model tier, with better reasoning and the latest tool-use behavior.
  • GPT-5.5 — if you need frontier quality rather than the cheapest option.

There’s no cost argument for staying on GPT-3.5: the recommended replacements are both cheaper and more capable. The only real work is the migration itself.

How to migrate off gpt-3.5-turbo

For most apps this is a small, low-risk change:

  1. Find every GPT-3.5 call. Grep your codebase and configs for gpt-3.5-turbo, including any pinned snapshots like -0125 or -1106 and any fine-tuned model IDs.
  2. Swap the model string. Replace it with your chosen successor (for example gpt-4o-mini or a GPT-5 mini model). The Chat Completions request shape is unchanged.
  3. Re-test prompts and outputs. Newer models follow instructions more literally and format differently; review prompts that depended on GPT-3.5’s quirks, and check any JSON or function-calling paths.
  4. Re-create fine-tunes. Fine-tuned gpt-3.5-turbo models retire with the base model — retrain on a current base if you rely on one.
  5. Update token and cost assumptions. Larger context windows and different pricing change your budgeting and rate-limit math.

Done early, this is an afternoon’s work. Left until October, it becomes an outage.

Building fast and short on engineers?

Need to modernize an AI feature or migrate a production integration without breaking it? Codersera connects you with vetted remote AI and backend engineers, fast.

Hire vetted remote developers →

FAQ

Is GPT-3.5 still available in 2026?

In the API, yes — for now. gpt-3.5-turbo still responds to calls but is scheduled to shut down on October 23, 2026. In the ChatGPT app it has been unavailable since July 2024.

When is gpt-3.5-turbo being shut down?

October 23, 2026 for gpt-3.5-turbo, gpt-3.5-turbo-1106, and fine-tuned variants. gpt-3.5-turbo-instruct shuts down on September 28, 2026. Always confirm against OpenAI’s official deprecations page.

What replaces GPT-3.5?

For the cheap-and-fast use case, GPT-4o mini and the GPT-5 mini / nano tier are the practical successors — both cheaper and more capable than GPT-3.5, with vision and much larger context windows.

Will migrating off GPT-3.5 break my app?

Usually not. The Chat Completions API shape is the same, so most migrations are a model-string swap plus prompt re-testing. The main exceptions are fine-tuned models (which must be retrained) and prompts tuned to GPT-3.5’s specific behavior.

Why did OpenAI retire GPT-3.5?

It has been superseded on every axis. Newer small models are cheaper and far more capable, so maintaining the older generation no longer makes sense — the same reason GPT-3.5 left ChatGPT in 2024.