Meta

Llama 4 Maverick

Generally availableOpen weights1M context

Summary

Llama 4 Maverick is an open-weight language model from Meta, released on 5 Apr 2025. It accepts text and images and generates text, with a 1M-token context window and up to 16K output tokens per response.

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...

Last verified 18 Nov 2026. Every figure on this page links to its source.

Specifications

DeveloperMeta
Release date5 Apr 2025 [OpenRouter]
StatusGenerally available
WeightsOpen weights [Hugging Face]
Licensellama4 [Hugging Face]
Parameters402B [Hugging Face]
Context window1M (1,048,576 tokens) [OpenRouter]
Max output16K tokens [OpenRouter]
Knowledge cutoff31 Aug 2024 [OpenRouter]
InputText and images [OpenRouter]
OutputText [OpenRouter]
Reasoning modeNo [OpenRouter]
Tool callingYes [OpenRouter]
Structured outputYes [OpenRouter]
Hugging Facemeta-llama/Llama-4-Maverick-17B-128E-Instruct [Hugging Face]

Llama 4 Maverick API pricing by provider

USD per million tokens. Sorted by input price.

We do not currently track a paid API listing for this model.

Running Llama 4 Maverick locally

Estimated memory, from the parameter count on Hugging Face. Includes about 0.8 GB of runtime overhead.

QuantizationWeightsTotal, 8K contextTotal, 32K context
Q4_K_M (4-bit)243 GB244 GB–
Q8_0 (8-bit)427 GB427 GB–
FP16 / BF16 (16-bit)803 GB804 GB–

At 4-bit with an 8K conversation, Llama 4 Maverick needs about 244 GB, more than any single consumer graphics card; it needs a unified-memory machine, a multi-GPU setup or a split between graphics card and system RAM. The model's architecture is not public, so the conversation cache is not included. Check your own hardware, or see Best computer for local AI on our sister site AIPCs.

Frequently asked questions

What is the context window of Llama 4 Maverick?

Llama 4 Maverick has a context window of 1,048,576 tokens (1M) and can generate up to 16,384 tokens in a single response, according to OpenRouter.

Is Llama 4 Maverick open source?

Llama 4 Maverick is an open-weight model: its weights are published on Hugging Face as meta-llama/Llama-4-Maverick-17B-128E-Instruct under the llama4 license. Check the license terms before commercial use.

When was Llama 4 Maverick released?

Meta released Llama 4 Maverick on 5 Apr 2025, according to OpenRouter.

How much VRAM does Llama 4 Maverick need?

About 244 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 427 GB at 8-bit. That is more than any single consumer graphics card holds. These are estimates from the model's published parameter count.

Similar models from other labs

Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.

ModelLabReleasedContextInput $/1MOutput $/1MWeights
Gemini 2.5 Pro Preview 06-05Google DeepMind5 Jun 20251M–––
MiniMax M1MiniMax17 Jun 20251M–––
MiniMax M1 80KMiniMax17 Jun 20251M$0.55$2.20Open
UnslopNemo 12BThedrummer8 Nov 20241M$0.40$0.40Open
Nova Premier 1.0Amazon31 Oct 20251M–––
DeepSeek ReasonerDeepSeek1 Dec 20251M––Open

Change history

  • No longer listed on openrouter
  • openrouter price changed: input $0.19 → $0.19, output $0.65 → $0.65 per 1M tokens

Sources