Groq Unveils LLaVA V1.5 7B, Faster than OpenAI GPT-4o

Groq has introduced LLaVA v1.5 7B, a new visual model now available on its Developer Console. This launch makes GroqCloud multimodal and broadens its support to include image, audio, and text modalities.

— Benjamin Klieger (@BenjaminKlieger) September 4, 2024

LLaVA, short for Large Language and Vision Assistant, combines language and vision capabilities. It builds on OpenAI’s CLIP and Meta’s Llama 2 7B model, utilising visual instruction tuning to enhance image-based natural instruction following and visual reasoning. 

This enables LLaVA to excel in tasks such as visual question answering, caption generation, optical character recognition, and multimodal dialogue.

“LLaVA-v1.5-7B which supports vision/image inputs and in our initial benchmarking response times were >4X faster than GPT-4o on OpenAI,” said Artificial Analysis.

— Artificial Analysis (@ArtificialAnlys) September 4, 2024

The new model unlocks numerous practical applications. Retailers can use it for inventory tracking, social media platforms can improve accessibility with image descriptions, and customer service chatbots can handle text and image-based interactions. 

Additionally, it aids in automating tasks in industries such as manufacturing, finance, retail, and education by streamlining processes and enhancing efficiency.

Developers and businesses can LLaVA v1.5 7B in Preview Mode on GroqCloud. 

Groq recently partnered with Meta, making the latest Llama 3.1 models—including 405B Instruct, 70B Instruct, and 8B Instruct—available to the community at Groq speed.

Former OpenAI researcher Andrej Karpathy praised Groq’s inference speed, saying, “This is so cool. It feels like AGI—you just talk to your computer and it does stuff instantly. Speed really makes AI so much more pleasing.”

Founded in 2016 by Ross, Groq distinguishes itself by eschewing GPUs in favour of its proprietary hardware, the LPU.