OpenAI Is Betting Everything on Gemini to Counter Gemini and Anthropic

Listen to This Article

Sam Altman freezes non-essential projects and doubles down on ChatGPT and a new Garlic model as Google’s Gemini 3 surges ahead.

OpenAI is moving from victory lap to survival mode. CEO Sam Altman has declared a company-wide code red, pausing a slate of new initiatives. This has fast-paced teams to focus on shoring up ChatGPT and preparing a new model, Garlic, aimed at Google’s Gemini 3 and Anthropic’s Claude Opus 4.5.

According to internal memos reported by The Information, The Wall Street Journal and others, Sam Altman told staff that OpenAI would delay advertising, shopping and health agents, and a personal assistant called Pulse. They will concentrate engineering effort on core ChatGPT upgrades like speed, reliability, and better personalization.

In one note, he warned employees that OpenAI is at “a critical time for ChatGPT” and should expect “rough vibes” and temporary economic headwinds. Additionally, OpenAI is still loss-making despite an estimated valuation of around $500 billion, with plans to spend up to $1.4 trillion on data centers and infrastructure by 2033. Internal projections suggest it may need roughly $200 billion in annual revenue to reach profitability by 2030.

Google’s Gemini 3 Flips the Script

Google’s Gemini 3 family starts to push OpenAI onto the back foot. The latest Gemini model has scored strongly across popular AI benchmarks and now sits at the heart of Google Search, the Gemini app and key Workspace tools.

Google’s ecosystem advantage is already visible in usage. After Google introduced its Nano Banana image generator, Gemini’s monthly active users reportedly climbed from 450 million in July to about 650 million by October, even as ChatGPT maintains an edge with some 800 million weekly users.

It is a reversal of 2022, when Google itself declared “code red” in response to ChatGPT’s launch. Now, OpenAI is the one trying to catch up as investors watch whether its huge infrastructure bets are justified.

Garlic: OpenAI’s new bet on reasoning and code

Garlic is a new large language model that OpenAI’s chief research officer Mark Chen recently described internally as performing “well” against both Gemini 3 and Anthropic’s Opus 4.5 in coding and reasoning tests.

Reports suggest Garlic builds on an earlier in-house project called Shallotpeat, fixing pre-training issues and packing “big model” knowledge into a more efficient architecture. In internal evaluations, Altman has told staff that OpenAI’s latest reasoning model is already “ahead” of Gemini 3, and the company is racing to ship Garlic as GPT-5.2 or GPT-5.5 as early as 2026.

For now, the message to staff is blunt: ChatGPT must get faster, smarter, and more reliable, and Garlic has to land strongly enough to reset the competitive narrative. Whether that is enough to cool OpenAI’s “code red” will depend on how quickly those promises show up in products that users actually stick with.