Llama 4 Maverick
Language modelOpen weightsCard summarizedOptimized for multimodal understanding, multilingual tasks, coding, tool-calling, and powering agentic systems
● Re-checked 12 Oct 2026 · 64 facts · 0 corrections · 1 unconfirmedSource: Llama 4 Model Card (MODEL_CARD.md)
Use it via API
Where to call it
- Self-host (weights)
meta-llama/Llama-4-Maverick-17B-128E-Instruct - llama.com downloads
Llama 4 Maverick
Good for
- Commercial and research use in multiple languages
- Instruction-tuned: assistant-like chat and visual reasoning
- Visual recognition, image reasoning, captioning and image Q&A
Overview
Llama 4 Maverick
Optimized for multimodal understanding, multilingual tasks, coding, tool-calling, and powering agentic systems
- Cost
- $0.24 host
- Context
- 1M
- Cheapest host
- $0.24 / $0.97
- Max output
- —
- Knows up to
- August 2024
- Released
- 5 Apr 2025
Llama 4 Maverick is a natively multimodal, mixture-of-experts language model (17B active / 400B total, 128 experts) from Meta's Llama 4 collection. It takes multilingual text and images and returns multilingual text and code. Maverick is the larger of the two released Llama 4 models and scores highest of the Llama models on Meta's reported benchmarks. Model Information
meta-llama/Llama-4-Maverick-17B-128E-Instruct- Input / output per 1M
- $0.24 / $0.97 via Amazon Bedrock
- Context window
- 1M
- Max output
- —
- Knows up to
- August 2024
- Reasoning
- —
- Weights
- Open · Llama 4 Community License
About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses Meta's Llama 4 Model Card (MODEL_CARD.md). Every claim carries its source page.