# Llama 4 Maverick

Meta, released 2025-04-05

**Verdict: conditional.** The leading open-source model for teams that need self-hosting, data sovereignty, or want to avoid API vendor lock-in. Last verified 2026-05-20.

## Pricing

- Input: Free (self-hosted) or $0.20 / 1M tokens (hosted)
- Output: Free (self-hosted) or $0.60 / 1M tokens (hosted)
- Context: 1M tokens

## Strengths

- Self-hosted inference
- Custom fine-tuning for domain-specific tasks
- Data sovereignty compliance
- Reducing API costs at scale
- Tasks requiring the highest quality reasoning
- Teams without GPU infrastructure
- Quick prototyping where API access is faster

## Benchmarks

- MMLU-Pro: 73.7% (https://ai.meta.com/blog/llama-4-multimodal-intelligence/)
- HumanEval: 78.6% (https://ai.meta.com/blog/llama-4-multimodal-intelligence/)
- LiveBench Reasoning: 76.6% (https://ai.meta.com/blog/llama-4-multimodal-intelligence/)

## Sources

- [Llama 4 Blog Post](https://ai.meta.com/blog/llama-4-multimodal-intelligence/)
- [Hugging Face: Llama-4-Maverick-17B-128E-Instruct model card](https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct)
- [Meta Llama](https://ai.meta.com/llama/)

Canonical: https://neomanex.com/models/llama-4-maverick