#United States #Mountain View #AI Inference #Groq #Llama 4

Groq Unveils Llama 4 Models on GroqCloud

In a significant advancement in the realm of artificial intelligence, Groq has officially launched Meta's Llama 4 models—Scout and Maverick—on their GroqCloud platform as of April 5, 2025. This initiative symbolizes a leap forward, providing developers and enterprises immediate, day-zero access to some of the most sophisticated open-source AI models currently available.

One of the standout features of this launch is that Groq has streamlined the entire process. By manufacturing their unique chips specifically for inference, Groq is capable of running these models without any delays, tuning, or performance bottlenecks. Moreover, they have achieved the remarkable feat of offering the lowest cost per token in the industry, which significantly enhances affordability for organizations wanting to implement AI solutions.

Cost Breakdown

With the launch of Llama 4 models, Groq presents an enticing pricing structure:

• - Llama 4 Scout: Accessing the Scout model costs $0.11 per million input tokens and $0.34 per million output tokens, averaging at $0.13 per million tokens overall.
• - Llama 4 Maverick: This model is priced at $0.50 per million input tokens and $0.77 per million output tokens, leading to a blended rate of $0.53 per million tokens.

These competitive rates enable developers to handle cutting-edge multimodal workloads economically, keeping costs manageable while ensuring low-latency responses.

About Llama 4 Models

Meta’s latest models, the Llama 4 family, utilize a Mixture of Experts (MoE) architecture that supports a native multimodal capacity. The two variants include:

• - Llama 4 Scout (17Bx16E): This general-purpose model excels in tasks such as summarization, reasoning, and code execution. It operates at a speed of over 460 tokens per second on Groq infrastructure.
• - Llama 4 Maverick (17Bx128E): Optimized for multilingual and multimodal applications, this model is particularly adept for creative applications and intelligences like chatbots, making it a powerful tool for industry professionals.

Quick Start with GroqCloud

Developers can commence their journey with Llama 4 on GroqCloud seamlessly. The models can be accessed through:

• - GroqChat
• - GroqCloud Developer Console
• - Groq API (with model IDs available in the console)

For those interested, free access is available, and users can upgrade their plans for enhanced rate limits and increased throughput to suit their respective needs. Developers are encouraged to start building right away at Groq Console.

About Groq

Groq stands at the forefront of AI inference technology, ensuring low-cost, high-performance systems without compromising on quality. Their cutting-edge LPU (Logic Processing Unit) and optimized cloud infrastructure reliably implement the most powerful open-source models instantly. Currently, Groq supports over a million developers, demonstrating their commitment to innovation and scalability in AI.

Conclusion

The arrival of Llama 4 models on GroqCloud heralds a new era in AI development, empowering enterprises with powerful tools at competitive pricing. As organizations continue to seek out efficient AI solutions, Groq’s offerings pave the way for enhanced productivity and innovation in various sectors—becoming a pivotal player in the AI landscape.