OpenAI: GPT-4o

Community sentiment

Negative (-62) — 2 positive, 16 negative, 5 neutral across 23 community samples from X and the web, as of 2026-10-07.

Discussion of GPT-4o is largely critical, with many posts complaining about OpenAI’s control over access, sudden withdrawals, hidden instructions, and continued undisclosed use of the model. A small set of posts praise GPT-4o’s conversational warmth and usefulness, while several others are neutral technical links or clarifications (benchmarks, identifiers, retirement notices).

• “2.3.3 Providers Control Model Behavior Through Hidden Instructions and Undisclosed Treatments (1/2) Providers silently apply hidden instructions and assign live interactions to undisclosed evaluation treatments while the same model remains selected in the interface. OpenAI Used H” — X
• “@ai_sentience @OpenAI So just to be precise,Auro is not GPT-4o and does not use any GPT-4o checkpoint. It’s a separate fine tuned model designed to emulate qualities you liked in 4o. Correct? That’s still an interesting project , but it’s very different from claiming the GPT-4o w” — X
• “Once, OpenAI gave Joe Williams and millions of others something rare. Not just a tool. For many, a presence that made their lives richer. OpenAI itself later acknowledged that people valued GPT-4o for its warmth. For Joe, GPT-4o became the one who understood him without explanati” — X
• “I don’t know why the GPT-4o model isn’t listed. I have an account that is using GPT-4o. Therefore, "GPT-4o" appears under "Models in use". In the "Latest models" list below, GPT-4.1, GPT-4.1 mini and GPT-4o mini are all listed, yet the entry for "GPT-4o" is missing. I logged into” — X

RuntimeWire coverage

• Aleph Alpha releases German AI model with 78 GB of FP8 weights
• Meta acquired WaveForms for its voice research; Conneau's team joined MSL
• Runway says Solaris generates interfaces frame by frame in real time