Llama 4 Maverick
Llama 4 Maverick is Meta's larger April 2025 open-weight model, a mixture-of-experts design with 128 experts and about 17B parameters active per token. It accepts text and images and was trained for 12 languages. Meta's newer releases use the Muse name, so Maverick is no longer the company's headline model.
Specifications
- Context window
- 1Mabout 786,000 words, or 1,570 pages
- Max output
- Not published
- Input price
- —no list price confirmed
- Output price
- —
- Cached input
- Not published
- Weights
- OpenLlama 4 Community License
- Parameters
- 400B total, 17B activeMoE
- Input types
- Text, Images
- Output types
- Text
- Reasoning mode
- No
- Tool calling
- Yes
- Released
- Apr 5, 2025Legacy
No serverless list price found on the official pricing pages of Together AI, Fireworks, DeepInfra or Groq as of 2026-10-10 (Together lists Llama 4 only for fine-tuning; Groq and DeepInfra do not list Maverick). Set to null.
What it costs in practice
| Workload | Monthly cost at list price |
|---|---|
| Support chatbot30,000 conversations a month, 2,000 input and 400 output tokens each | No list price |
| Document summaries2,000 documents a month, 12,000 input and 800 output tokens each | No list price |
| Coding agent1,000 tasks a month, 60,000 input and 6,000 output tokens each | No list price |
Standard rates, no caching or batch discounts. Run your own numbers.
Strengths
- Native image and text input in one model
- 1M-token context window
- FP8 weights fit on a single H100 DGX host
Trade-offs to weigh
- About 400B total parameters must sit in memory even though only 17B are active
- Llama 4 Community License adds attribution rules and a 700M monthly-user threshold
Where it fits
Meta positions Llama 4 Maverick for:
Where you can use it
- Hugging Face
- Ollama
Run it yourself
Released under Llama 4 Community License. Approximate memory needed just to load the weights:
| Precision | Memory for weights |
|---|---|
| 16-bit (BF16/FP16) | 800 GB |
| 8-bit (FP8/INT8) | 400 GB |
| 4-bit (Q4_K_M, about 4.85 bits) | 242 GB |
This is a mixture-of-experts model: all 400B total parameters must sit in memory, but only about 17B are used per token, so it runs faster than a dense model of the same size.
Add memory for the KV cache (it grows with context length and concurrent users) and 1-2 GB of runtime overhead. Size it precisely in the VRAM calculator.
Alternatives to consider
| Model | Tier | Context | Input / output per 1M | |
|---|---|---|---|---|
| Mistral Large 3Mistral AI | Advanced | 256K | $0.50 / $1.50 | Compare |
| Mistral Medium 3.5Mistral AI | Advanced | 256K | $1.50 / $7.50 | Compare |
| DeepSeek-V4.1-FlashDeepSeek | Advanced | 1M | $0.30 / $1.20 | Compare |
| gpt-oss-120bOpenAI | Advanced | 128K | $0.15 / $0.60 | Compare |
Frequently asked questions
How much does Llama 4 Maverick cost?
We couldn't confirm a published per-token list price for Llama 4 Maverick on October 10, 2026. No serverless list price found on the official pricing pages of Together AI, Fireworks, DeepInfra or Groq as of 2026-10-10 (Together lists Llama 4 only for fine-tuning; Groq and DeepInfra do not list Maverick). Set to null. Check the provider's pricing page for current rates.
What is the context window of Llama 4 Maverick?
Llama 4 Maverick accepts up to 1,048,576 tokens of context (about 786,000 words, or 1,570 pages). Long prompts cost more and can be slower, so most apps send only what the task needs.
Is Llama 4 Maverick open source?
Llama 4 Maverick is an open-weight model released under the Llama 4 Community License license. You can download the weights from Hugging Face (meta-llama/Llama-4-Maverick-17B-128E-Instruct) and run or fine-tune them yourself, within the license terms. Open weights isn't the same as fully open source: training data and code may not be published.
Can I run Llama 4 Maverick locally?
Yes, if you have enough memory. The weights alone need about 242 GB at 4-bit quantization and about 800 GB at 16-bit, plus room for the KV cache. Use the VRAM calculator on LLModel.com to check a specific GPU or Mac.
What is Llama 4 Maverick best for?
Meta positions Llama 4 Maverick for multimodal, multilingual, long documents, self-hosting. It accepts text, images as input, doesn't offer a separate reasoning mode and supports tool calling. Test it on a sample of your own tasks before you choose it for production.
Sources
- huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct
- huggingface.co/unsloth/Llama-4-Maverick-17B-128E-Instruct/raw/main/config.json
- dev.meta.ai/llama/docs/model-cards-and-prompt-formats/llama4
- www.together.ai/pricing
- deepinfra.com/pricing
- console.groq.com/docs/models
- ollama.com/library/llama4
Facts checked against these pages on October 10, 2026. Spot an error? Tell us.
Still choosing a model?
Send us your use case and constraints. We'll come back with a shortlist and an evaluation plan you can run on your own data.
- Free and vendor-neutral: no model provider pays for placement.
- Your details go only to partners you agree to.