Llama 4 Maverick VRAM Requirements: GPU Needed to Run It

Llama 4 Maverick needs about 212.9 GB at Q4_K_M with 8K context.

Llama 4 Maverick is the wide mixture: 128 local experts and 1 active per token, across 48 layers. The parameter total includes the vision tower. Unlike Scout, this file does not scale context.

The maximum is 1,048,576 tokens, and the table itself stops at 131,072. Chunked layers still use an 8,192-token chunk. Q4_K_M at 8,192 tokens is 212.9 GB.

No single GPU in our list fits this; see multi-GPU or CPU offload. The download is gated and the public architecture copy is Unsloth. There is no packed 4-bit expert format and no compressed latent cache.

Query and key head counts match Scout. The expert fan-out does not.

Which GPUs fit

FP16 Q8_0 Q6_K Q5_K_M Q4_K_M Q3_K_M Q2_K

Fit at FP16/BF16 and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	738.5 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	742.5 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	740.5 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	726.5 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	742.5 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	738.5 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	726.5 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	738.5 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · Won't fit	32 GB	718.5 GB short	Won't fit
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	726.5 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	734.5 GB short	Won't fit
NVIDIA A100 80GB80 GB · Won't fit	80 GB	670.5 GB short	Won't fit
NVIDIA H100 SXM 80GB80 GB · Won't fit	80 GB	670.5 GB short	Won't fit
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · Won't fit	48 GB	714.5 GB short	Won't fit

Fit at Q8_0 and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	387.9 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	391.9 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	389.9 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	375.9 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	391.9 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	387.9 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	375.9 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	387.9 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · Won't fit	32 GB	367.9 GB short	Won't fit
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	375.9 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	383.9 GB short	Won't fit
NVIDIA A100 80GB80 GB · Won't fit	80 GB	319.9 GB short	Won't fit
NVIDIA H100 SXM 80GB80 GB · Won't fit	80 GB	319.9 GB short	Won't fit
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · Won't fit	48 GB	363.9 GB short	Won't fit

Fit at Q6_K and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	297.3 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	301.3 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	299.3 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	285.3 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	301.3 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	297.3 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	285.3 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	297.3 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · Won't fit	32 GB	277.3 GB short	Won't fit
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	285.3 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	293.3 GB short	Won't fit
NVIDIA A100 80GB80 GB · CPU offload	80 GB	229.3 GB short	CPU offload
NVIDIA H100 SXM 80GB80 GB · CPU offload	80 GB	229.3 GB short	CPU offload
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · Won't fit	48 GB	273.3 GB short	Won't fit

Fit at Q5_K_M and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	247.6 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	251.6 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	249.6 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	235.6 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	251.6 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	247.6 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	235.6 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	247.6 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · Won't fit	32 GB	227.6 GB short	Won't fit
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	235.6 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	243.6 GB short	Won't fit
NVIDIA A100 80GB80 GB · CPU offload	80 GB	179.6 GB short	CPU offload
NVIDIA H100 SXM 80GB80 GB · CPU offload	80 GB	179.6 GB short	CPU offload
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · Won't fit	48 GB	223.6 GB short	Won't fit

Fit at Q4_K_M and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	200.9 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	204.9 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	202.9 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	188.9 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	204.9 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	200.9 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	188.9 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	200.9 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · Won't fit	32 GB	180.9 GB short	Won't fit
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	188.9 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	196.9 GB short	Won't fit
NVIDIA A100 80GB80 GB · CPU offload	80 GB	132.9 GB short	CPU offload
NVIDIA H100 SXM 80GB80 GB · CPU offload	80 GB	132.9 GB short	CPU offload
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · Won't fit	48 GB	176.9 GB short	Won't fit

Fit at Q3_K_M and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	151.2 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	155.2 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	153.2 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	139.2 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	155.2 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	151.2 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	139.2 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	151.2 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · Won't fit	32 GB	131.2 GB short	Won't fit
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	139.2 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	147.2 GB short	Won't fit
NVIDIA A100 80GB80 GB · CPU offload	80 GB	83.20 GB short	CPU offload
NVIDIA H100 SXM 80GB80 GB · CPU offload	80 GB	83.20 GB short	CPU offload
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · Won't fit	48 GB	127.2 GB short	Won't fit

Fit at Q2_K and 8,192 tokens. Cards that fit in usable memory are listed first. "Short" shows how far the model goes over usable memory. Headroom uses usable memory. A unified row uses 75 percent of the listed pool.
GPU	Memory	Status	Headroom
GeForce RTX 3060 12GB12 GB · Won't fit	12 GB	113.2 GB short	Won't fit
GeForce RTX 3070 8GB8 GB · Won't fit	8 GB	117.2 GB short	Won't fit
GeForce RTX 3080 10GB10 GB · Won't fit	10 GB	115.2 GB short	Won't fit
GeForce RTX 3090 24GB24 GB · Won't fit	24 GB	101.2 GB short	Won't fit
GeForce RTX 4060 8GB8 GB · Won't fit	8 GB	117.2 GB short	Won't fit
GeForce RTX 4060 Ti 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
GeForce RTX 4070 12GB12 GB · Won't fit	12 GB	113.2 GB short	Won't fit
GeForce RTX 4070 Ti Super 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
GeForce RTX 4080 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
GeForce RTX 4090 24GB24 GB · Won't fit	24 GB	101.2 GB short	Won't fit
GeForce RTX 5060 Ti 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
GeForce RTX 5070 12GB12 GB · Won't fit	12 GB	113.2 GB short	Won't fit
GeForce RTX 5070 Ti 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
GeForce RTX 5080 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
GeForce RTX 5090 32GB32 GB · CPU offload	32 GB	93.22 GB short	CPU offload
Radeon RX 7900 XTX 24GB24 GB · Won't fit	24 GB	101.2 GB short	Won't fit
Radeon RX 9070 XT 16GB16 GB · Won't fit	16 GB	109.2 GB short	Won't fit
NVIDIA A100 80GB80 GB · CPU offload	80 GB	45.22 GB short	CPU offload
NVIDIA H100 SXM 80GB80 GB · CPU offload	80 GB	45.22 GB short	CPU offload
Apple M5 Max 48GB unified48 GB · 36.00 GB usable · CPU offload	48 GB	89.22 GB short	CPU offload

VRAM by quantization

Totals at 8,192 tokens. Q4_K_M is the usual balance of size and quality. FP16 weights are the parameter count times 2 bytes. The FP16 column in the table is the full estimate: weights, cache, and 1 GB.

Estimates at 8192 tokens, in GB. Formula rows apply one block size to every parameter. A published-file row uses measured weight bytes. FP16 weights equal the parameter count times 2 bytes.
Quantization	Bits	Weights	Cache	Overhead	Total
FP16/BF16Bits 16 · Cache 1.50 · Overhead 1.00	16	748.0	1.50	1.00	750.5
Q8_0Bits 8.5 · Cache 1.50 · Overhead 1.00	8.5	397.4	1.50	1.00	399.9
Q6_KBits 6.5625 · Cache 1.50 · Overhead 1.00	6.5625	306.8	1.50	1.00	309.3
Q5_K_MBits 5.5 · Cache 1.50 · Overhead 1.00	5.5	257.1	1.50	1.00	259.6
Q4_K_MBits 4.5 · Cache 1.50 · Overhead 1.00	4.5	210.4	1.50	1.00	212.9
Q3_K_MBits 3.4375 · Cache 1.50 · Overhead 1.00	3.4375	160.7	1.50	1.00	163.2
Q2_KBits 2.625 · Cache 1.50 · Overhead 1.00	2.625	122.7	1.50	1.00	125.2

Context length

Q4_K_M total as the context changes. Cache is the part that grows.

2,048 211.8 GB
8,192 212.9 GB
32,768 214.0 GB
131,072 218.5 GB

Cache and totals as the context changes. Tokens above the configured maximum are not listed.
Context tokens	Cache GB	Q4_K_M total	FP16 total
2048FP16 total 749.4	0.38	211.8	749.4
8192FP16 total 750.5	1.50	212.9	750.5
32768FP16 total 751.6	2.63	214.0	751.6
131072FP16 total 756.1	7.13	218.5	756.1

The configured context length is 1,048,576 tokens. The context table stops at 131,072 tokens.

Llama 4 Maverick hardware notes

Unified memory counts 75 percent of the listed pool as usable. Discrete cards use the full listed memory. The memory column still shows the listed figure.
Last updated 2025-05-22.
Weights, cache, and the 1 GB overhead are estimates. How this is calculated.
Architecture numbers come from unsloth/Llama-4-Maverick-17B-128E-Instruct because the official download is gated. The parameter count is the official Hugging Face record.

Llama 4 Maverick VRAM FAQ

How much VRAM does Llama 4 Maverick need at Q4_K_M?

212.9 GB at 8,192 tokens. Weights are 210.4 GB, the cache is 1.50 GB, and overhead is 1 GB.

Does a 24 GB GPU fit Llama 4 Maverick at Q4_K_M?

GeForce RTX 3090 24GB is Won't fit, 188.9 GB over. The quant table uses 8,192 tokens, the smaller of 8,192 and the configured maximum of 1,048,576.

What context length is configured for Llama 4 Maverick?

The configured context length is 1,048,576 tokens. The quant table uses 8,192 tokens, the smaller of 8,192 and the configured maximum of 1,048,576.

Which source supplies the parameter count for Llama 4 Maverick?

401.6B parameters, from the Hugging Face model record for meta-llama/Llama-4-Maverick-17B-128E-Instruct. Architecture numbers come from the public copy at unsloth/Llama-4-Maverick-17B-128E-Instruct, because the official download is gated. The parameter count is the official record.

Models in the same family, then the closest Q4_K_M totals.
Model	Q4_K_M GB
DeepSeek R1	360.1
DeepSeek V3	360.1
gpt-oss-120b	62.49
Llama 3.1 70B	40.46

Llama 4 Maverick VRAM and GPU Requirements