Sun Apr 06 09:40:00 UTC 2025: ## Meta’s Llama 4 Models Challenge ChatGPT, But Fall Short in Some Areas

**SAN FRANCISCO** – Meta has released two new large language models (LLMs), Llama 4 Maverick and Llama 4 Scout, entering the AI chatbot arena already dominated by OpenAI’s ChatGPT. While Meta boasts Llama 4 Maverick’s superior performance in coding, reasoning, and multilingual tasks based on benchmark tests like LMarena, a direct comparison reveals a mixed bag.

Llama 4 Maverick currently leads GPT-4o and GPT-4.5 Preview on LMarena, trailing only Gemini 2.5 Pro. However, Meta’s claims of excellence in creative writing and image generation lack independent verification. Furthermore, its multimodal capabilities (handling image prompts) are currently limited to English and US users. In contrast, ChatGPT’s recently unveiled native image generation feature has impressed users with its photorealistic and contextually accurate outputs, including image editing capabilities. Tests reveal ChatGPT’s image generation significantly surpasses that of Meta’s current offering.

A key difference is the absence of a dedicated reasoning model in Meta’s current lineup. Meta plans to address this with a Llama 4 Reasoning model launching later this month. Moreover, ChatGPT’s “Deep Research” mode, providing enhanced web search capabilities, is currently unavailable in Llama 4. While Llama 4 Maverick offers unrestricted free access via Meta’s platforms (WhatsApp, Instagram, Messenger), ChatGPT limits free usage, including image generation. Meta also plans to release a more powerful model, Llama 4 Behemoth, to compete directly with GPT-4.5 and Claude Sonnet 3.7.

In conclusion, while Meta’s Llama 4 models represent a significant advancement, ChatGPT maintains an edge in certain features, particularly image generation and deep web search. The competitive landscape is rapidly evolving, with both companies promising further advancements in the near future.

Read More