gpt-4-turbo retirement date 2026-10-23
official substitute gpt-5.6-sol
As of 2026-10-07, the 2026-04-22 section of OpenAI's deprecations page still lists October 23, 2026 as the shutdown date for gpt-4-turbo (the same cell names gpt-4-turbo-2024-04-09 and gpt-4-turbo-completions), with gpt-5.6-sol as the substitute model. The same day gpt-4 / gpt-4-0613 move to gpt-5.6-sol and gpt-3.5-turbo / gpt-3.5-turbo-0125 to gpt-5.6-terra. Both substitutes can be called through QCode.
Updated 2026-10-07
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four separate facts
Chat-wave shutdown day
Official table, 2026-04-22 section: October 23, 2026 | gpt-4-turbo | gpt-4-turbo-2024-04-09, gpt-4-turbo-completions | gpt-5.6-sol (checked 2026-10-07).
Replacement for turbo / gpt-4
gpt-4 and gpt-4-turbo rows point at gpt-5.6-sol. gpt-3.5-turbo that day points at gpt-5.6-terra. Do not remap every old id to one name.
gpt-4o-2024-05-13 and ft-gpt-4
gpt-4o-2024-05-13 also shuts down on October 23, 2026, substitute gpt-5.6-sol. The fine-tuned entry ft-gpt-4 is removed the same day with gpt-5.6-sol as the recommended base model; ft-gpt-3.5-turbo maps to gpt-5.6-terra.
When resellers 404
The table governs OpenAI's own API. When QCode starts rejecting old ids is not predicted here; plan to finish before 2026-10-23.
How to handle dated chat-id shutdowns
As of 2026-10-07 the method is unchanged: grep gpt-4-turbo, gpt-4, gpt-3.5-turbo and their aliases, follow the Substitute model column row by row, and do not remap everything to Sol from memory. Then send one tiny request and confirm the echoed model is the new id. Through QCode you can call the substitutes the table names, gpt-5.6-sol and gpt-5.6-terra; whether an old id still answers after the shutdown is not promised here.
What the 2026-10-23 section looks like
The deprecations page fetched on 2026-10-07, section 2026-04-22: Legacy GPT model snapshots, says “Access to these models will be shut down on the dates below.” Its October 23, 2026 chat rows: gpt-4-turbo (same cell: gpt-4-turbo-2024-04-09, gpt-4-turbo-completions) → gpt-5.6-sol; gpt-4-0613 (gpt-4, gpt-4-0613-completions, gpt-4-completions) → gpt-5.6-sol; gpt-3.5-turbo-0125 (gpt-3.5-turbo, gpt-3.5-turbo-completions) → gpt-5.6-terra; gpt-4o-2024-05-13 → gpt-5.6-sol. Next on the schedule: on 2026-12-11 older GPT-5 and o3 snapshots such as gpt-5-2025-08-07 and o3-2025-04-16 shut down, both onto gpt-5.6-sol (notice of 2026-06-11). On 2027-04-01 gpt-5.3-codex and gpt-5.1 go to gpt-6-sol and gpt-5.4-nano to gpt-6-luna, announced 2026-10-01 with six months' notice.
Timeline
The 2026-04-22: Legacy GPT model snapshots section of the table; this wave's shutdown date reads October 23, 2026.
Re-checked 2026-10-07: the gpt-4-turbo, gpt-4 and gpt-3.5-turbo rows are still there, dated October 23, 2026; the section now sits under the page's Past deprecations heading while its text still says will be shut down.
Shutdown date on the table. OpenAI states no postponement. Production should finish the swap before this day.
Three evidence tiers
Officially announced / confirmed
Verifiable word for word on 2026-10-07: the 2026-04-22 section has the columns Shutdown date / Model snapshot / Substitute model; October 23, 2026 shuts gpt-4-turbo and the aliases in its cell, substitute gpt-5.6-sol; gpt-4-0613 / gpt-4 → gpt-5.6-sol; gpt-3.5-turbo-0125 / gpt-3.5-turbo → gpt-5.6-terra; ft-gpt-4 → gpt-5.6-sol and ft-gpt-3.5-turbo → gpt-5.6-terra. OpenAI's definition: at the time of the shut down, the model or endpoint will no longer be accessible.
Unpublished / caution
First, on the 2026-10-07 page this section is listed under the Past deprecations heading, yet the table still reads October 23, 2026 and the text still says will be shut down; OpenAI gives no reason for the placement, so this page reads it neither as a shutdown that has happened nor as a postponement. Second, there is no postponement wording; OpenAI only says some developers may be able to provision dedicated capacity after the shutdown date through sales, which is not a general delay. Third, when QCode rejects old ids is unpublished and not predicted here. Fourth, the table maps ids only; whether a substitute behaves the same is yours to test.
turbo → Sol versus 3.5 → terra
gpt-4-turbo / gpt-4 → Sol
The table names gpt-5.6-sol for gpt-4-turbo, gpt-4 and gpt-4o-2024-05-13. On QCode billing is per token; unit prices for each model are on /models.
gpt-3.5-turbo → terra
A different row the same day. Do not silently remap 3.5-turbo to Sol.
Three steps before the shutdown date
① Search: grep code, environment variables and SDK defaults for gpt-4-turbo, gpt-4-turbo-2024-04-09, gpt-4-turbo-completions, gpt-4, gpt-4-0613, gpt-3.5-turbo and gpt-3.5-turbo-0125, aliases and snapshots together. ② Change: set model per the substitute column, gpt-4-turbo and gpt-4 to gpt-5.6-sol, gpt-3.5-turbo to gpt-5.6-terra. ③ Verify: send one non-streaming tiny request and confirm the response reports the new model id.
What to do on QCode
gpt-5.6-sol and gpt-5.6-terra can be called through QCode, billed per token; unit prices for each model are on /models. Per the QCode docs page on endpoints and API formats: the OpenAI Python/JS SDK base_url is https://api.qcode.cc/openai/v1 (requests go to /openai/v1/chat/completions); Codex CLI uses base_url https://api.qcode.cc/openai (requests go to /openai/v1/responses). GPT models use these two OpenAI endpoints only, not the Anthropic endpoint. After switching model from gpt-4-turbo to gpt-5.6-sol, send one non-streaming tiny request and check the echoed model.
FAQ
Is 10-23 officially confirmed?
Yes. As of 2026-10-07 the deprecations table still prints October 23, 2026 and the substitute column.
Will they delay?
The table carries no postponement wording (checked 2026-10-07). OpenAI mentions that some developers may provision dedicated capacity after the shutdown date through sales; that is not a delay. Plan for 2026-10-23.
Remap every old GPT to Sol?
No. gpt-3.5-turbo’s row is terra. Read the table.
Same as the 09-28 instruct wave?
No. 09-28 is Completions-era ids. 10-23 is the chat wave.
When will QCode 404 old ids?
Unpublished. Finish before the official day.
Does gpt-4o-2024-05-13 also shut down on 10-23?
Yes. It sits in the same October 23, 2026 table with gpt-5.6-sol as substitute. The fine-tuned entries ft-gpt-4 (→ gpt-5.6-sol) and ft-gpt-3.5-turbo (→ gpt-5.6-terra) are removed the same day.
Sources
Shutdown dates and substitutes: OpenAI deprecations page https://developers.openai.com/api/docs/deprecations (sections 2026-04-22, 2026-06-11 and 2026-10-01, fetched 2026-10-07). Integration settings: the QCode docs page on endpoints and API formats at docs.qcode.cc (fetched 2026-10-07). OpenAI writes dates as October 23, 2026; this page renders them as ISO.
Move gpt-4-turbo to gpt-5.6-sol before 2026-10-23
One key for gpt-5.6-sol and gpt-5.6-terra, billed per token, unit prices on /models. Whether a retired id stays callable is not promised here.
Related
OpenAI legacy model retirement dates
The 09-28 batch’s shutdown date has passed (instruct → terra); overview of the 10-23 batch.
Sol vs Sonnet 5
Daily comparison after you land on Sol.
OpenAI API triage
A shutdown 404 is not a transient 5xx.
Not affiliated with OpenAI. Checked 2026-10-07; dates and substitutes follow OpenAI's official deprecations page, and model availability follows /models.