Even as CEO Sam Altman is busy redirecting all internal resources to augment ChatGPT, reports claim that the AI powerhouse is working on a new large language model (LLM). The new LLM codenamed ‘Garlic’ reportedly outperforms rivals in use cases such as coding and reasoning tasks.

The development was reported by The Information and seems to come at a time when the AI startup is facing intense pressure from its peers like Google and Anthropic, which recently introduced their state-of-the-art AI models. Reportedly, the new model signals the company’s strategic move into specialised high-value industries such as biomedicine and healthcare.

Also, the ‘Garlic’ model indicates a shift from generalist AI to specialised AI applications. Mark Chen, chief research officer at OpenAI, told his coworkers that the model has been performing well in internal evaluations when compared to Google’s Gemini 3 and Anthropic’s Claude Opus 4. 5, especially in coding and reasoning capabilities.

OpenAI is expected to introduce Garlic as GPT-5. 2 or GPT-5. 5 by early 2026.

Competition intensifies Reports about Garlic coincides with Altman’s declaration of ‘Code Red’ on Monday, December 1. The CEO in an internal memo urged employees to prioritise improvements to ChatGPT.

At the same time, he asked them to delay other initiatives, including the company’s wider advertising plans. The OpenAI boss said that the company would steer more resources towards augmenting ChatGPT’s responsiveness and personalisation features as competition intensifies. The sense of urgency by OpenAI is palpable as the AI landscape is undergoing a dramatic shift with Google releasing Gemini 3 on November 18 and integrating it across its ecosystem.

Gemini 3 is reportedly Google’s fastest-ever rollout, and the model has demonstrated top performance on benchmark leaderboards for image editing, text generation, and multimodal reasoning. Similarly, Anthropic also released its Claude Opus 4. 5 last month, claiming it to be the world’s best model for coding.

Perhaps, the major reason for OpenAI’s concern is its dwindling traffic. According to Deedy Das, an entrepreneur from Palo Alto, California, in the two weeks since the launch of Gemini 3, ChatGPT’s unique daily active users (7-day average) went down by -6 per cent.

OpenAI relies on its consumer business for its revenue, and this is propelled by strong user growth. Das, in his LinkedIn post, said that OpenAI’s lofty $500B valuation relies on continued user momentum to meet its year-end target of $20B ARR. This is seemingly the first big setback in the company’s history.

Story continues below this ad Gemini’s rapid growth Another report in Fortune claimed that Google’s Gemini app has grown to 650 million monthly active users in October, up from 450 million in July. Interestingly, three years ago Google had declared ‘Code Red’ in response to ChatGPT’s launch and its subsequent widespread growth.

Also Read | OpenAI and Google impose new daily limits for Sora, Nano Banana AI tools amid demand surge Reportedly, Altman’s memo also hinted that the company will release a new model next week which is ahead of Gemini 3. The memo indicates that the company is focused on enhancing the model behaviour and image generation capabilities. It is to be noted that Google’s Nano Banana Pro gained traction owing to its image generation capabilities.

Chen, who took the chief research officer role earlier this year, has led major OpenAI projects including DALL·E, Codex and o1 reasoning models. Even as some senior researchers have departed for rivals this year, OpenAI still retains a large research team that the company will likely lean on to close the competitive gap.