Llama 4 Maverick
by Meta
Llama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active
Models
Llama 4 Maverick
Added 10 July 2026
Overview
Llama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active parameters per forward...
Best for
Best for
Teams that want open-weight flexibility with hosted convenience
Use cases
- self-host experiments
- cost-controlled apps
- fine-tune pipelines
How to use / API access
Call via OpenRouter with model id `meta-llama/llama-4-maverick`.
- openrouter ·
meta-llama/llama-4-maverickDocs
OpenRouter id: meta-llama/llama-4-maverick
Benchmarks
- Intelligence Index (artificial-analysis) : 14.3 index
- Coding Index (artificial-analysis) : 16.3 index
- Agentic Index (artificial-analysis) : 1.3 index
- Design Arena 3d (other) : 966 elo · #97
- Design Arena codecategories (other) : 916 elo · #107
- Design Arena dataviz (other) : 918 elo · #106
- Design Arena gamedev (other) : 896 elo · #108
Enterprise DNA angle
EDNA helps teams decide when open weights belong in Omni vs when a closed frontier model is cheaper in total cost of quality.
Notes
Llama 4 Maverick is listed in the EDNA Models directory from the OpenRouter catalogue.
Pros
- Open-weights family economics
- Strong community and self-host path
Cons
- Hosted quality depends on the OpenRouter endpoint
- May trail closed frontier models on hardest tasks
Get the free Developer’s Field Guide
A 27-page field guide to the AI coding workflow with Claude. Claude Code, MCP servers, the prompt patterns that work, and what to delegate. Free.
Enter your work email. We send it straight over, plus a few short notes worth knowing. Unsubscribe any time.