💬 Start a Chat with this Setup
Credits / Run
10
Model Type
LLM – Mixture of Experts (LLaMA 4)
Connection
Replicate API
Parallel Agents
Up to 3 simultaneous sessions

LLaMA 4 Maverick – 17B MoE Instruction Model

Overview:

LLaMA 4 Maverick is a cutting-edge 17 billion parameter instruction model featuring 128 expert modules and a mixture-of-experts architecture. Built for high-performance natural language understanding and generation, this model delivers efficient, high-quality completions and excels at both general-purpose prompts and task-specific instructions. It supports nuanced responses, follows user intent with high fidelity, and is optimized for low-latency, real-world applications.

Features & Highlights

  • ⚡ 17B parameters with 128 expert modules
  • ⚡ Mixture-of-experts architecture for dynamic routing
  • ⚡ Instruction-tuned for task-specific use
  • ⚡ Efficient token generation with strong coherence
  • ⚡ Real-time chat capabilities with API and UI support