<!-- Lazu · Gemini 2.5 Pro retired on Lazu: old names end Nov 15 · https://lazu.ai/blog/model-retirement-aliases-2026-11 -->

# Gemini 2.5 Pro retired on Lazu: old names end Nov 15

2026-11-03 · Lazu team

*Compiled by Molly · Lazu team*

Short answer: on November 3, 2026, gemini-2.5-pro and 16 other legacy models were retired from Lazu. The old names keep working until November 15, 2026: until then, requests that use them do not fail; they are answered by the matching new model in the table below and billed at that model's price. After November 15 the old names are deleted and requests using them fail. Please change the model names in your code before then.

## What happened

| Date | Event |
| --- | --- |
| 2026-09-18 | Google limits the Gemini 2.5 models to users who had actively used them before; new projects cannot call them |
| 2026-10-23 | Lazu's upstream supply for gemini-2.5-pro starts returning 429, intermittently |
| 2026-11-01 | Almost every request fails |
| 2026-11-03 | Google deprecates gemini-2.5-pro and gemini-2.5-flash and announces that from December 1 API requests are routed to gemini-3.8-flash and gemini-3.5-flash-lite; the same day Lazu retires gemini-2.5-pro and 16 other legacy models |
| 2026-11-15 | Old model names are deleted; requests using them fail |

The root cause is supply. After September 18 only existing accounts could reach 2.5 Pro, so every third party's 2.5 Pro capacity shrank, and by late October our route no longer got through. Keeping the model listed would mean selling a model known to fail, so it goes.

## Why the old names last until November 15

Model names live in code, environment variables and agent configs. Delete a name the day a model is retired and old code starts failing with "model not found" at three in the morning. So there is a short transition of under two weeks:

- the catalog no longer shows these models, so new users cannot pick them;
- until November 15, requests with an old name are still accepted and answered by the matching new model;
- during that time they are billed at the new model's price, and usage details show that price;
- after November 15 the old names are deleted and requests fail.

The transition is there so nothing breaks overnight; please don't rely on it. We have emailed everyone with an address.

## Old to new

| Family | Retired model | Switch to |
| --- | --- | --- |
| Gemini | gemini-2.5-pro | gemini-3.1-pro-preview |
| Gemini | gemini-3-flash-preview, gemini-3.5-flash | gemini-3.6-flash |
| GPT | gpt-5 | gpt-5.6-sol |
| GPT | gpt-5-mini | gpt-5.6-terra |
| GPT | gpt-5-nano | gpt-5.6-luna |
| GPT | gpt-5.4-nano | gpt-6-luna |
| GPT | gpt-5.1 | gpt-6-sol |
| Claude | claude-haiku-4-5-20251001 | claude-haiku-5-5 |
| Claude | claude-sonnet-4-6 | claude-sonnet-5-5 |
| Claude | claude-opus-4-6, claude-opus-4-7, claude-opus-5 | claude-opus-5-5 |
| Claude | claude-fable-5 | claude-fable-5-1 |

gemini-3.1-pro-preview-customtools and the two dated snapshot names gpt-5.4-nano-2026-03-17 and gpt-5.4-mini-2026-03-17 are retired the same way.

The GPT rows follow OpenAI's own schedule: gpt-5, gpt-5-mini and gpt-5-nano shut down on December 11, 2026, and the gpt-5.6 family is OpenAI's named replacement. For Claude, Anthropic's earliest retirement date for claude-haiku-4-5 was October 15, 2026, which has passed.

## Why we suggest 3.1 Pro and not Google's own 3.8 Flash

Google's deprecations page names gemini-3.8-flash as the replacement for gemini-2.5-pro, and from December 1 its API routes requests that way. We did not copy that. The reason is tier: 2.5 Pro was the Pro tier of its generation and 3.8 Flash is a Flash model; reasoning depth and price sit in different bands. gemini-3.1-pro-preview is the only Pro-named Gemini model Google still serves, and Google's own error message for 2.5 Pro says to use it. People calling 2.5 Pro wanted Pro, so we point them to Pro.

One caveat: gemini-3.1-pro-preview is still a preview. Google has announced no shutdown date and no replacement for it. Google's current flagship is gemini-4-argon, which joined the catalog in this update.

## What to change in your code

1. Search your code, environment variables and agent configs for the names in the left column and replace them with the right column.
2. If you budget or alert by model name, recompute with the new name and price.
3. Finish before November 15, 2026. After that the old names are deleted and requests fail.

## Three models added the same day

- gemini-4-argon: Google's flagship reasoning model, 1M shared context.
- gpt-6.1-terra: OpenAI's efficient model for visual reasoning and computer use.
- gemini-nano-banana-2.1: Google's October 2026 update to Nano Banana 2, image generation and editing at 1K, 2K and 4K.

Prices are on each model's page.

## Upcoming shutdowns by the labs

| Date | Model | On Lazu |
| --- | --- | --- |
| 2026-12-01 | Google routes gemini-2.5-pro and gemini-2.5-flash requests to newer models | gemini-2.5-pro already retired; gemini-2.5-flash was never listed |
| 2026-12-11 | OpenAI shuts down gpt-5, gpt-5-mini, gpt-5-nano | Already retired; switch to the gpt-5.6 family |
| 2027-01-14 | OpenAI shuts down gpt-4o-mini, gpt-4.1, gpt-4.1-mini | Will be retired, with email notice first |

## Related

- [Gemini 3.8 Flash doubles in price on Jan 1: how to budget](/blog/gemini-3-8-flash-price-2027)
- [GPT-6.1 Sol vs GPT-6 Sol](/blog/gpt-6-1-sol-vs-gpt-6-sol)

