Best GPU for Llama 4 Maverick: VRAM Requirements
Llama 4 Maverick needs about 881.8 GB of GPU memory in BF16 and 221.8 GB at 4 bits. See which GPUs fit it for inference and fine-tuning.
Llama 4 Maverick needs about 881.8 GB of VRAM in FP16, 441.8 GB in FP8 and 221.8 GB at 4 bits for inference with an 8,192-token context. In FP16 it fits on 4 × NVIDIA B300 288GB; a 4-bit build fits on one NVIDIA B300 288GB.
Llama 4 Maverick is a mixture-of-experts language model with 400B total and 17B active parameters per token from Meta, released in April 2025, with a 1,048,576-token context window and published under the Llama 4 Community License. Its weights alone take 800 GB in BF16 and 200 GB at 4 bits per parameter; the tables below add the KV cache and runtime memory, then list the GPUs that hold each setup.
Recommended GPUs for Llama 4 Maverick
Best for inference
2 GPUs × NVIDIA B300 288GB
FP8 or INT8 weights, 441.8 GB at 8,192 tokens.
Best for fine-tuning
8 GPUs × NVIDIA B200 180GB
LoRA on BF16 weights, 1,040 GB.
Budget pick
4 GPUs × NVIDIA RTX PRO 6000 Blackwell 96GB
Smallest card class that holds a 4-bit build, 221.8 GB.
Llama 4 Maverick VRAM requirements by precision
One request with an 8,192-token context, KV cache in BF16, plus a 10 % margin for activations and runtime buffers.
| Precision | Weights | KV cache | Total for inference |
|---|---|---|---|
| BF16 / FP16 | 800 GB | 1.6 GB | 881.8 GB |
| FP8 | 400 GB | 1.6 GB | 441.8 GB |
| INT4 | 200 GB | 1.6 GB | 221.8 GB |
Llama 4 Maverick VRAM by task
Fine-tuning rows assume a short sequence, batch 1 and gradient checkpointing; longer sequences add activation memory.
| Task | GPU memory |
|---|---|
| Inference at the full 1,048,576-token context (FP8) | 666.8 GB |
| QLoRA (4-bit base) | 260 GB |
| LoRA (BF16 base) | 1,040 GB |
| Full fine-tuning (BF16, AdamW) | 7,040 GB |
GPUs that fit Llama 4 Maverick
Number of cards whose combined memory holds each setup, assuming the model is split across them with tensor or pipeline parallelism. A dash means more than eight cards.
| GPU | Memory | BF16 inference (881.8 GB) | FP8 inference (441.8 GB) | INT4 inference (221.8 GB) | QLoRA (260 GB) | LoRA (1,040 GB) |
|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 32GB | 32 GB | — | — | 8 GPUs | — | — |
| NVIDIA Tesla V100 32GB | 32 GB | — | — | 8 GPUs | — | — |
| NVIDIA A100 PCIe 40GB | 40 GB | — | — | 8 GPUs | 8 GPUs | — |
| NVIDIA L40S 48GB | 48 GB | — | — | 8 GPUs | 8 GPUs | — |
| NVIDIA Quadro RTX 8000 48GB | 48 GB | — | — | 8 GPUs | 8 GPUs | — |
| NVIDIA RTX 6000 Ada 48GB | 48 GB | — | — | 8 GPUs | 8 GPUs | — |
| NVIDIA RTX A6000 48GB | 48 GB | — | — | 8 GPUs | 8 GPUs | — |
| NVIDIA A100 PCIe 80GB | 80 GB | — | 8 GPUs | 4 GPUs | 4 GPUs | — |
| NVIDIA A100 SXM 80GB | 80 GB | — | 8 GPUs | 4 GPUs | 4 GPUs | — |
| NVIDIA H100 PCIe 80GB | 80 GB | — | 8 GPUs | 4 GPUs | 4 GPUs | — |
| NVIDIA H100 SXM 80GB | 80 GB | — | 8 GPUs | 4 GPUs | 4 GPUs | — |
| NVIDIA RTX PRO 6000 Blackwell 96GB | 96 GB | — | 8 GPUs | 4 GPUs | 4 GPUs | — |
| NVIDIA H200 141GB | 141 GB | 8 GPUs | 4 GPUs | 2 GPUs | 2 GPUs | 8 GPUs |
| NVIDIA B200 180GB | 180 GB | 8 GPUs | 4 GPUs | 2 GPUs | 2 GPUs | 8 GPUs |
| NVIDIA B300 288GB | 288 GB | 4 GPUs | 2 GPUs | 1 GPU | 1 GPU | 4 GPUs |
| NVIDIA GB300 NVL | 288 GB | 4 GPUs | 2 GPUs | 1 GPU | 1 GPU | 4 GPUs |
What is Llama 4 Maverick?
Llama 4 Maverick is the larger Llama 4 mixture-of-experts model from Meta, with 128 experts and 17 billion active parameters out of 400 billion. It reads text and images, writes text in 12 languages and supports a 1-million-token context window.
- Developer
- Meta
- Released
- April 2025
- Parameters
- 400B total, 17B active per token
- Context window
- 1,048,576 tokens
- License
- Llama 4 Community License
Three of every four layers use chunked attention over 8,192 tokens, so the real KV cache at long context is far below this estimate. Meta also publishes FP8 weights that fit on a single 8-GPU H100 server.
Source: Llama 4 Maverick 17B 128E Instruct model card (as of November 17, 2026).
Frequently asked questions
- How much VRAM does Llama 4 Maverick need?
- For inference with an 8,192-token context and one request, Llama 4 Maverick needs about 881.8 GB in BF16, 441.8 GB in FP8, 221.8 GB in INT4, including a 10 % runtime margin. Real usage depends on the inference engine and its settings.
- Can Llama 4 Maverick run on a single GPU?
- Yes. In INT4 it needs about 221.8 GB, which fits one NVIDIA B300 288GB (288 GB). Higher precisions need a larger card or several GPUs.
- How much GPU memory does fine-tuning Llama 4 Maverick take?
- QLoRA (4-bit base): about 260 GB; LoRA (BF16 base): about 1,040 GB; Full fine-tuning (BF16, AdamW): about 7,040 GB. Longer training sequences and larger batches add activation memory on top.
- What is the best GPU for Llama 4 Maverick?
- Best for inference: 2 GPUs × NVIDIA B300 288GB (FP8 or INT8 weights, 441.8 GB at 8,192 tokens). Best for fine-tuning: 8 GPUs × NVIDIA B200 180GB (LoRA on BF16 weights, 1,040 GB). Budget pick: 4 GPUs × NVIDIA RTX PRO 6000 Blackwell 96GB (Smallest card class that holds a 4-bit build, 221.8 GB). Check live offers before booking, as availability changes.
- Can Llama 4 Maverick run on an RTX 4090?
- No: even a 4-bit build needs about 221.8 GB, more than eight RTX 4090 cards hold (192 GB). It needs data center GPUs with more memory per card.
- Can Llama 4 Maverick run on an RTX 5090?
- Not on one card: even a 4-bit build needs about 221.8 GB, so it takes 8 RTX 5090 cards (256 GB combined) with the model split across them.
Estimates use the formulas of the LLM VRAM calculator (decimal GB). Real usage depends on the inference engine, quantization format and memory fragmentation: test on the target GPU before a long rental.
Explore the live market
Every offer on the live market comes from a reviewed supplier and is sold by EverXCloud Inc.