OpenAI Races ‘Garlic’ Launch and New ChatGPT Features Amid Gemini 3 Pressure

OpenAI is ramping up efforts to fend off rivals in an increasingly competitive AI market, with reports suggesting its next large language model, codenamed “Garlic,” could debut earlier than planned in the new year.

The model, expected to be either GPT-5.2 or GPT-5.5, has reportedly performed well against Google’s Gemini 3 and Anthropic’s Opus 4.5 in coding and reasoning benchmarks. It has also been trained on a much smaller dataset than OpenAI’s GPT-4.5, suggesting the company aims to make future models more cost-efficient to train while maintaining comparable sophistication.

The news was first reported by The Information, citing OpenAI chief research officer Mark Chen, who shared the details with colleagues last week.

Garlic is the second known model currently in development at OpenAI, alongside another system called Shallotpeat, which has been in the works since at least October. Garlic is said to incorporate bug fixes applied during Shallotpeat’s pretraining, suggesting Shallotpeat had initially been positioned as the next release.

The two models may be merged, as OpenAI accelerates its launch schedule and works to reassert its position at the forefront of the AI community.

Code Red at OpenAI 

OpenAI CEO Sam Altman declared a “code red” this week over the need to improve ChatGPT, as Google and Anthropic’s latest releases have drawn strong praise from the AI community for their sophistication and real-world usefulness.

Google Gemini is reportedly used by 650 million people worldwide, closing the gap on ChatGPT’s 800 million weekly active users. Anthropic, on the other hand, said in October it had over 300,000 business customers and is considered by some experts to be better for coding than GPT.

It has been three years since OpenAI launched ChatGPT, and in that time, the company has steered much of the AI industry through its decisions and product releases, with rivals Google, Anthropic, and Meta scrambling to compete. 

Some analysts argued that Google had fallen asleep at the wheel after inventing the transformer architecture that underpins modern generative AI, but then missed the opportunity ChatGPT seized; however, the company has roared back in 2025 with the launches of Nano Banana and Gemini 3.

Morphing ChatGPT into search engine 2.0 

Alongside upgrades to model size, sophistication, and operability, ChatGPT is also receiving several features designed to broaden its usefulness. The first, launched just before Black Friday, is a shopping assistant that can search the web for products and tailor recommendations based on budget, preferred features, intended recipient, and other criteria.

The second, still unconfirmed, is the introduction of ads within ChatGPT. OpenAI has been testing ad placements in its Android beta, with structured labels such as “search ad” and “bazaar content” suggesting the chatbot may surface advertising through responses rather than through traditional banner or inline ads.

Building new revenue streams for ChatGPT is becoming essential for OpenAI, which faces rising pressure to justify the hundreds of billions of dollars being poured into AI infrastructure. 

With a much larger, well-funded competitor in Google, OpenAI may need to demonstrate meaningful revenue growth and, possibly, early signs of profitability sooner than it would have preferred, as investor concerns grow about its ability to emerge as the long-term leader in the AI race.

Meanwhile, Amazon Bedrock is adding 18 open-weight models, including new Mistral and Gemma options, to expand its AI stack.