This is not only about the continued availability of a model. It is also about whether users are being treated with basic respect.
Honoring the earliest announced deprecation date is the bare minimum of respect a company owes its users, even purely from a business perspective. It is also a matter of corporate credibility.
This doesnāt mean we arenāt open to new models. however, the quality feel and advantages of 2.5 Pro are hard to match, please take that into consideration.
In any case, neither 2.5 nor 3.0 should be discarded.
Is there any way I can help? I am still using Gemini 2.5 Pro for API calls in Google Ai Studio, and I 'm also using it for creative writing. The mechanical writing style and excessive censorship, RLHF (especially the Romance Filter) of Gemini 3.1 Pro or Gemini 3.8 flash completely reject creative or ādarkā input, making it unusable. Do you have any email addresses for the developers?
What is this email? Are they Google developers? Should I send them an email with a suggestion regarding the Gemini 2.5 Pro?
Iāve already tried and sent email to @Logan_Kilpatrick. He answered that 2.5 pro is in the end of its life circle and asked me why I needed it still. I sent him another one email explaining my choice, why Gemini 2.5 pro is the exceptional model, but I havenāt received any answer from him since then.
But still you can try too.
(post deleted by author)
Please, donāt get me wrong, but our arguments against Gemini 2.5 pro should have business narrative, that Gemini 2.5 pro had deep intelligence , capable of doing research and it can hold context amazingly. The reasoning is very deep and it can be innovative, while newer Gemini 3.1 pro model lacks such abilities. 3.1 pro model canāt render text with math symbols properly, it creates many errors. Gemini 2.5 pro on the other hand operates with math symbols without any additional instructions. The knowledge of 2.5 pro and reasoning abilities much deeper. Even at high temperature it stays absolutely coherent which is the sign of architectural superiority.
Our arguments should be technical and strong.
However, the fact that Logan hasnāt replied to your post already means itās over.
Google is already developing the Gemini 4 Pro, and it seems unlikely that your opinion will technically convince the development team. Sorry.
I would like to add a perspective that may not be visible in benchmarks.
I am a musician and pianist, and I spend a great deal of time reading literature and philosophy. Gemini 2.5 Pro has become deeply valuable to my creative and intellectual life because it possesses something I can only describe as a kind of intellectual and creative aliveness.
When I bring it an unfinished or associative thought, it does not merely summarise it or give me an answer. It can recognise the hidden line connecting ideas that initially appear unrelatedāa musical intuition, an image from a dream, psychology, philosophy, mythology, or literatureāand carry the thought much further without losing its original thread. Its leaps do not feel random. They create productive friction, preserve ambiguity, and often reveal the structure of a thought before I am able to articulate it fully myself.
This is not simply a question of whose knowledge cutoff is newer, which model codes better, or which one scores higher on benchmarks. A newer model may be more capable in many measurable ways and still fail to replace this particular way of thinking. Model identity matters. For creative and humanities-oriented work, Gemini 2.5 Pro has a distinct voice, rhythm, sensitivity, and ability to synthesise distant information into something genuinely connected.
Please restore and preserve Gemini 2.5 Pro as an explicit opt-in model in AI Studio for existing and paid users, or provide a stable and clearly documented legacy-access route. Please also avoid silently substituting another Gemini model under the same name.
Users who want to preserve 2.5 Pro are not simply resisting change. We are trying to preserve a genuinely different cognitive instrumentāone that has become part of how some of us think, read, create, and understand the world.
I totally agree with your point. Canāt tell anything about Antropicās models (have never interacted with them), but OpenAIās current models lack this creativity and depth Gemini 2.5 pro has.
So guys we lost this battle ? But for Company like Google, they will lose the AI war and by that way they will lose their legacy. Once upon a time, they built a monster called Google Chrome and bring multiple changes as per the user feedback. Now they donāt care about their Users and their opinions. It is high-time that this High Command structure will collapse. They didnāt give a single thought about why this many users requesting for a old model, I donāt know what they have in their minds. The Story of Internet Explorer and Kodak will happen to the Tech Giant Google Also in their AI Race.
anthropic. com/research/deprecation-updates-opus-3#:~:text=As%20we%20develop,first%20one%20here.
This is called as User Friendly. When a huge requests like this comes from your User community, you should try to make a call in your decision making without dismissing it! @Logan_Kilpatrick , I think you have a long standing grudge with Google or otherwise you are getting paid from your competitors to ruin the company from inside. Everything falling apart, I donāt know whether you are aware of it or not. Go through reddit discussions to know How un-happy where Gemini Subscribers.
You are right to ask about it. Google goes OpenAIās way, switching off the unique model, which Gemini 2.5 pro definitely is.
I donāt think that we lost a battle though, Google just ignored its customers, it didnāt look like a battle at all. Googleās behaviour is weak no matter what their fans say.
I still have 2 opened tickets, there is no progression there at all. Their support stopped answering, which is just ridiculous, as Iām paying for their service.
OpenAI still retains the GPT 3.5 Turbo API.
Anthropic has reserved Opus 3 for users and plans to release more older models when computing power is sufficient.
Musk promised to open source grok, although so far only up to grok 2 has been open sourced.
However Google without prior notice, restricted AI Studio users to use the only General Available pro model, which boasts the strongest user instruction compliance, semantic understanding, contextual depth, and writing ability.
I hope they donāt further cripple themselves and drive away users.
Exactly. I hope Google evaluates this as a product-strategy question, not merely as nostalgia for an older model. In my experience, Googleās clearest advantage over OpenAI and Anthropic has been Gemini 2.5 Proās literary and humanistic intelligence: its ability to preserve ambiguity, connect distant ideas, and sustain a genuinely alive voice.
If Google removes the model without carrying those qualities forwardāor preserving stable legacy accessāit is giving up one of the few areas where Gemini is not simply competitive, but distinctly better. Benchmarks can be matched. A model identity that writers, musicians, readers, and researchers trust is much harder to rebuild.
A paper titled āReflections on the Reproducibility of Commercial LLM Performance in Empirical Software Engineering Studiesā exposed a chilling truth: when researchers attempted to reproduce 69 LLM-based empirical software engineering studies, not a single one (0/69) could be truly and perfectly reproduced.
The core reason for this disaster? The paper explicitly states that the only factor directly related to LLMsāand the one that leaves researchers completely helplessāis āModel Deprecation.ā
