Aug. 11, 2012 - Seventy-two hours after a public revolt by its most loyal usersOpenAI Forced to dramatically adjust their strategy.
OpenAI CEO Aultman announced in an X post on Sunday that the company will beRecovery includes GPT-4o old version AI Modelsand significantly increaseUsage credits for paid subscribers, releasing goodwill to the angry user community.
Altman said early this morning that the percentage of daily use of inferential modeling has risen significantly:Free users increased from less than 1% to 7% and Plus users from 7% to 24%The increase in the call cap is therefore imperative.
On August 7, OpenAI launched GPT-5ChatGPT Plus is a new version of ChatGPT Plus, positioned as a "unified system" that automatically matches the best models, and removes the ability for users to select older models, such as GPT-4o. For ChatGPT Plus users, who pay $20 per month (note: current exchange rate is about $143.8), this is a cut-off of a workflow they've relied on for a long time:The different models have their own roles to playSome are good at creativity, some are good at logical reasoning, and some are long on in-depth research.
Harmonization to a single new model not only disrupted these arrangements, but also deprived users of a means to cross-verify their answers. Strong dissatisfaction quickly evolved into a wave of unsubscribes and online petitions.
According to Gizmodo, as public opinion heated up, Altman began to reassure users, admitting an error in judgment and announcing concessions. He responded directly to one user's question, "Will 4o be restored?" He responded directly to one user's question, "Will 4o be restored?" by saying, "It has been restored! Go to settings and select 'show old models'."
There was also a query about whether using the old version would be "labeled as weird," to which Altman responded, "I don't know if it's a good idea.Not in either case!”
To further assuage the discontent, Altman declaredSignificantly increase the invocation limit of the GPT-5 reasoning functionand said that all model categories will soon have higher caps than before GPT-5 was introduced. When asked about the exact value, he revealed "Now it's 3,000 times a week.”.
Altman also promises more transparency, with plans to interfaceDisplays the specific model of the current response requestThe company also released a blog post this week explaining the company's considerations for capacity allocation. He also released data showing that the average daily use of inference models by Plus users has risen from 7% to 24%, indirectly defending the company's initial strategy of pushing new technology.
In the end, OpenAI responded positively to user protests: retaining GPT-5 as the default while restoring the old option.