Faced with Google's offensive, OpenAI was in an uproar.
OpenAI issued a "red alert" to all employees:
We need to focus all our resources on getting ChatGPT to catch up with Gemini 3 Pro first, and then we can think about other things. Even things like advertising and shopping agents, which can bring in extra revenue, have to take a backseat .
Nano Banana's drawing is so good, it really should raise an alarm.
Three years ago, it was ChatGPT that triggered a red alert within Google, prompting the integration of Google Brain and DeepMind resources, which ultimately led to the creation of the Gemini series.
Now, in response to the Gemini 3 Pro, two new OpenAI models have been revealed:
- A new inference model is planned for release next week , which outperforms Gemini 3 in OpenAI's internal evaluation, but "more effort is still needed to improve the ChatGPT experience."
- The new model, codenamed Garlic , is being rushed to its initial version and may be released early next year as GPT-5.2 or GPT-5.5.
In addition to model improvements, Altman detailed a series of key areas that urgently needed improvement in an internal memo:
- Personalized services for over 800 million weekly active users
- Allows everyone to customize how they interact with ChatGPT.
- Improved image generation allows users to create a variety of content, from interior design models to animated photographs.
- Improve model behavior performance
- Surpassing competitors in public rankings such as LMAArena
- Improve the speed and reliability of ChatGPT
- Reduce "over-rejection," that is, situations where chatbots refuse to answer benign questions.
...
It's clear that they're in a real hurry.
For OpenAI employees, the already intense work pace can only become even more intense.
ChatGPT traffic plummets; OpenAI hunter becomes prey.
Why is OpenAI in such a hurry? The data speaks for itself.
Since the release of Geimin 3 Pro and Nano Banana Pro, ChatGPT traffic has dropped by 6% in 7 days.
According to data released by Google in October, Gemini's monthly active users have reached 650 million, compared to only 450 million just three months earlier in July.
What makes OpenAI even more nervous is that Google's newly released Gemini 3 model has surpassed OpenAI's model in industry benchmark tests, directly driving up Google's market value.
Furthermore, Google has launched an AI mode in its search application, transforming traditional search into a chatbot similar to ChatGPT. Meanwhile, Anthropic is gaining popularity in the enterprise market. Under this multi-pronged attack, OpenAI's leading advantage is rapidly shrinking.
Nick Turley, head of ChatGPT, revealed that ChatGPT currently accounts for 70% of global AI assistant activity and 10% of search activity. While the numbers still look impressive, signs of slowing growth are already emerging.
CFO Sarah Friar hinted at a slowdown in ChatGPT’s growth during a conference call with investors last month, although she did not specify which growth metric she was referring to.
ChatGPT's slowing growth means that its financial pressure is increasing day by day.
OpenAI is currently not profitable and needs to raise funds at an almost frantic pace to keep operating.
According to the company's own financial projections, ChatGPT will generate approximately $10 billion in revenue from its subscription services this year, with an estimated $20 billion next year and approximately $35 billion by 2027.
However, this revenue is far from covering expenses. OpenAI expects to burn through tens of billions of dollars over the next few years to develop new technologies and support products like ChatGPT.
Even more critically, the company needs to raise approximately $100 billion to cope with the massive cash burn. According to internal financial projections, OpenAI needs to grow its revenue to approximately $200 billion by 2030 to achieve profitability.
This financial disadvantage puts OpenAI at a disadvantage in competition with tech giants like Google, which can fund their investments with their own revenue, while OpenAI relies entirely on external financing.
Even compared to its main startup competitor Anthropic, OpenAI's spending is more aggressive.
Secret Weapon: New Model Codenamed "Garlic You're Ruthless"
OpenAI was not unprepared for the crisis.
According to sources, Chief Research Officer Mark Chen revealed a new model codenamed "Garlic" to some colleagues last week. In OpenAI's internal evaluation, this model outperformed Google's Gemini 3 and Anthropic's Opus 4.5 in both coding and inference tasks.
Garlic's model is not the same as the one previously revealed as "Shallotpeat" (tentative Chinese translation: mud is no good, shallot is scallion, and pearl is mud).
Shallotpeat is the model that Altman told employees in October would be used to challenge the Gemini 3. Garlic integrated bug fixes accumulated during the development of Shallotpeat during the pre-training process.
Mark Chen stated that significant breakthroughs were achieved in pre-training during the development of Garlic, enabling the packaging of knowledge from large models into smaller architectures.
OpenAI is releasing a version of Garlic as soon as possible, which means that people may see the release of GPT-5.2 or GPT-5.5 early next year.
Meanwhile, thanks to the lessons learned from Garlic, OpenAI has begun developing a larger and better model.
In addition, a model to compete with the Gemini 3 Pro will be released next week, making a total of 3 new models, which makes the already messy product line even more chaotic.
This red alert did not come suddenly. Memos show that OpenAI had previously declared an orange alert for improvements to ChatGPT.
Altman encouraged the mobilization of temporary teams and organized daily conference calls for those responsible for improving ChatGPT.
CFO Sarah Friar stated at an event in November that an IPO was not in the near future. However, as a privately held company, OpenAI's fate is closely tied to those of companies like Nvidia, Microsoft, and Oracle, and its commitment to investing hundreds of billions of dollars in data centers has raised concerns about its timeline for converting those investments into meaningful revenue.
The battle for AI is entering a fierce stage. ChatGPT, which caused panic for Google three years ago, is now experiencing the taste of being overtaken.
Whether OpenAI can hold its ground this time, and whether Garlic can help it widen the gap again, the entire tech world is holding its breath.
Reference link:
[1]https://www.wsj.com/tech/ai/openais-altman-declares-code-red-to-improve-chatgpt-as-google-threatens-ai-lead-7faf5ea6
[2]https://x.com/deedydas/status/1995879560446181633
This article is from the WeChat public account "Quantum Bit" , author: Meng Chen, and published with authorization from 36Kr.
