EXAONE 4.0 32B — how much card memory it needs

EXAONE 4.0 · 32B parameters · Q4_K_M · last updated 2026-09-26

Whether a quantised model file fits entirely in a graphics card's memory. It is not a statement about whether the model will run, how fast it will run, or how good its output will be.

These figures are calculated from published file sizes and model architecture. No combination on this site has been tested on physical hardware.

How much card memory it needs

Every figure below is either measured from the distributed file or calculated from the model architecture; the third column says which.
Component	Size	Where the figure comes from
Quantised file (Q4_K_M)	18.0 GiB	Measured: the size of the file actually distributed (19,343,826,720 bytes) · A read 2026-09-24 [S1]
Key/value cache at 4,096 tokens	1.0 GiB	Calculated from the model architecture under the assumptions below
Key/value cache at 32,768 tokens	2.8 GiB	Calculated from the model architecture under the assumptions below
Runtime overhead	1.0 GiB	Assumed: a flat allowance for compute buffers, the driver context and the display
Required at 4,096 tokens	20.1 GiB	Sum of the above
Required at 32,768 tokens	21.8 GiB	Sum of the above
Weights and overhead only	19.1 GiB	The floor, if the key/value cache is set aside

The arithmetic, so you can redo it

GiB (1024^3 bytes). Card memory advertised as '8GB' is treated as 8 GiB so the two are comparable.

required_vram_gib = quantised_file_size + kv_cache + runtime_overhead
kv_cache_bytes = 2 (key and value) * 2 bytes * num_kv_heads * head_dim * sum(tokens held per layer)
tokens_held_full_attention = num_layers * context_tokens
tokens_held_with_sliding_window = (num_layers - sliding_window_layers) * context_tokens + sliding_window_layers * min(context_tokens, sliding_window_tokens)

The inputs the formulas above take for this model.
Input for this model	Value
Layers	64 · A read 2026-09-24 [S2]
Key/value heads	8 · A read 2026-09-24 [S2]
Head dimension	128 · A read 2026-09-24 [S2]
Sliding window	4,096 tokens over 48 layers · A read 2026-09-24 [S2] · A read 2026-09-24 [S2]
Context length the model publishes	131,072 tokens
Distributed file	EXAONE-4.0-32B-Q4_K_M.gguf · A read 2026-09-24 [S1]
Distributor	LG AI Research (official), LGAI-EXAONE/EXAONE-4.0-32B-GGUF

Each graded figure above cites: [S1] LG AI Research official EXAONE-4.0-32B-GGUF distribution repository (file listing and actual file sizes); [S2] LG AI Research official EXAONE-4.0-32B config.json (num_hidden_layers=64, num_key_value_heads=8, head_dim=128, sliding_window=4096, sliding_window_pattern="LLLG" — three local layers to one global layer, repeating; layer_types breaks down as 48 sliding-attention layers and 16 full-attention layers).

Measured

The size in bytes of the quantised file that is actually distributed.
Model architecture: layer count, key/value head count, head dimension, sliding-window width.
Card memory size as published by the graphics vendor.

Assumed

The key/value cache is held at 16-bit precision for every token.
Two context lengths are assumed: 4,096 and 32,768 tokens.
A flat 1.0 GiB is set aside for compute buffers, the driver context and the display.

Every outcome here is calculated rather than stated by a vendor, so none of them is better than grade B. Where a figure an outcome rests on — the card's memory size, the size of the distributed model file, or the model's attention geometry — carries a weaker grade than that, the weaker grade is shown on the row alongside B, and the row says which figure it is.

Graphics cards with room for it

Card memory against this model's requirement. Nothing here was run on hardware.
Card	Card memory	Outcome	The arithmetic	Confidence
NVIDIA RTX PRO 6000 Blackwell Workstation Edition	96 GB	✓ Fits in VRAM	96 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA A100 80GB PCIe	80 GB	✓ Fits in VRAM	80 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA H100 PCIe	80 GB	✓ Fits in VRAM	80 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA A40	48 GB	✓ Fits in VRAM	48 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA L40	48 GB	✓ Fits in VRAM	48 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA L40S	48 GB	✓ Fits in VRAM	48 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA Quadro RTX 8000	48 GB	✓ Fits in VRAM	48 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX 6000 Ada Generation	48 GB	✓ Fits in VRAM	48 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX A6000	48 GB	✓ Fits in VRAM	48 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA A100 40GB PCIe	40 GB	✓ Fits in VRAM	40 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
AMD Radeon Pro W6800	32 GB	✓ Fits in VRAM	32 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
GeForce RTX 5090 (Reference design)	32 GB	✓ Fits in VRAM	32 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX 5000 Ada Generation	32 GB	✓ Fits in VRAM	32 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX PRO 4500 Blackwell	32 GB	✓ Fits in VRAM	32 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA Tesla V100 PCIe 32GB	32 GB	✓ Fits in VRAM	32 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
AMD Radeon RX 7900 XTX (Reference design)	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
GeForce RTX 3090 (Reference design)	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
GeForce RTX 3090 Ti (Reference design)	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
GeForce RTX 4090 (Reference design)	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA A10	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA A30	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA L4	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA Quadro M6000 24GB (Reference design)	24 GB	✓ Fits in VRAM Estimated	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA Quadro P6000 (Reference design)	24 GB	✓ Fits in VRAM Estimated	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA Quadro RTX 6000	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX 3090	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX 3090TI	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX 4500 Ada Generation	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX A5000	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX A5500	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA RTX PRO 4000 Blackwell	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA TITAN RTX (Reference design)	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
NVIDIA Tesla M40 24GB	24 GB	✓ Fits in VRAM Estimated	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
NVIDIA Tesla P40	24 GB	✓ Fits in VRAM	24 GB of memory is at least the 21.8 GB required at a context length of 32,768 tokens.	B
AMD Radeon RX 550X Discrete GDDR5	Not published	? Unverified	The memory size of this graphics option is not stated in the cited specification, so no verdict is given.	B
AMD Radeon RX 6500XT Discrete	Not published	? Unverified	The memory size of this graphics option is not stated in the cited specification, so no verdict is given.	B
GeForce GTX 1660 Discrete [gtx-1660-6gb]	Not published	? Unverified	The memory size of this graphics option is not stated in the cited specification, so no verdict is given.	B
GeForce GTX 1660 Discrete [gtx-1660-super-6gb]	Not published	? Unverified	The memory size of this graphics option is not stated in the cited specification, so no verdict is given.	B
GeForce RTX 3060 Discrete GDDR6	Not published	? Unverified	The memory size of this graphics option is not stated in the cited specification, so no verdict is given.	B
GeForce RTX 3070 Discrete GDDR6X	Not published	? Unverified	The memory size of this graphics option is not stated in the cited specification, so no verdict is given.	B
AMD Radeon RX 7900 XT (Reference design)	20 GB	✕ Exceeds VRAM	20 GB of memory is below the 20.1 GB required at a context length of 4,096 tokens.	B
NVIDIA RTX 4000 Ada Generation	20 GB	✕ Exceeds VRAM	20 GB of memory is below the 20.1 GB required at a context length of 4,096 tokens.	B
NVIDIA RTX 4000 SFF Ada Generation	20 GB	✕ Exceeds VRAM	20 GB of memory is below the 20.1 GB required at a context length of 4,096 tokens.	B
NVIDIA RTX A4500	20 GB	✕ Exceeds VRAM	20 GB of memory is below the 20.1 GB required at a context length of 4,096 tokens.	B
AMD Radeon RX 6800 (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6800 XT (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6900 XT	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6900 XT (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6950 XT (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 7600 XT (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 7800 XT (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 7900 GRE (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 9060 XT (Reference design)	16 GB	✕ Exceeds VRAM Estimated	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 9070 (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 9070 XT (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon VII (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon Vega Frontier Edition (Reference design)	16 GB	✕ Exceeds VRAM Estimated	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
GeForce RTX 4060 Ti (Reference design), 16 GB	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4070 Ti SUPER (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4080 (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4080 SUPER (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5060 Ti (Reference design), 16 GB	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5070 Ti (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5080 (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc A770 (Reference design), 16 GB	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA A16	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P5000 (Reference design)	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro RTX 5000	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX 2000 Ada Generation	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX A4000	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA T4	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Tesla P100 PCIe 16GB	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Tesla V100 PCIe 16GB	16 GB	✕ Exceeds VRAM	16 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6700 XT (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6750 XT (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 7700 XT (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX TITAN X (Reference design)	12 GB	✕ Exceeds VRAM Estimated	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce RTX 2060 12GB (Reference design)	12 GB	✕ Exceeds VRAM Estimated	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce RTX 3080 (Reference design), 12 GB	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3080 Ti (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4070 (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4070 SUPER (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4070 Ti (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5070 (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc B580 (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro M6000 (Reference design)	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX 3080TI	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX A2000 12GB	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA TITAN V (Reference design)	12 GB	✕ Exceeds VRAM Estimated	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA TITAN Xp (Reference design)	12 GB	✕ Exceeds VRAM Estimated	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
PALIT GeForce RTX 3060 Infinity 2 OC	12 GB	✕ Exceeds VRAM	12 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 1080 Ti (Reference design)	11 GB	✕ Exceeds VRAM Estimated	11 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce RTX 2080 Ti (Reference design)	11 GB	✕ Exceeds VRAM	11 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6700 (Reference design)	10 GB	✕ Exceeds VRAM	10 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3080 (Reference design), 10 GB	10 GB	✕ Exceeds VRAM	10 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc B570 (Reference design)	10 GB	✕ Exceeds VRAM	10 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX 3080	10 GB	✕ Exceeds VRAM	10 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon Pro W6600	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon R9 390 (Reference design)	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 480 (Reference design)	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 5500 XT (Reference design), 8 GB	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 5700 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 5700 XT (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 580 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 590 (Reference design)	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 6600 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6600 LE Discrete 8GB GDDR6	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6600 XT (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6650 XT (Reference design)	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 7600 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 9050 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX Vega 56 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX Vega 64 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 1070 (Reference design)	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1070 Ti (Reference design)	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1080 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 2060 SUPER (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 2070 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 2070 SUPER (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 2080 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 2080 SUPER (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3050 8GB (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3060 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3060 Ti (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3070 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3070 Ti (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4060 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4060 OC Low Profile 8G (GV-N4060OC-8GL)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 4060 Ti (Reference design), 8 GB	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5050 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5060 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5060 Ti (Reference design), 8 GB	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 5060 Ti 8G VENTUS 2X PLUS	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc A580 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc A750 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc A770 (Reference design), 8 GB	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA GeForce RTX 3050 Discrete 8GB GDDR6	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA GeForce RTX 4060 Discrete 8G GDDR6	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA GeForce RTX 5050 Discrete 8GB GDDR6	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA GeForce RTX 5060 Discrete 8G GDDR7	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA Quadro M4000 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro M5000 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P4000 (Reference design)	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro RTX 4000	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX A1000	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA T1000	8 GB	✕ Exceeds VRAM	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Tesla P4	8 GB	✕ Exceeds VRAM Estimated	8 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
AMD Radeon RX 5600 XT (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1060 (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1660 (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1660 SUPER (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1660 Ti (Reference design)	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 980 Ti (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX TITAN (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX TITAN Black (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce RTX 2060 (Reference design)	6 GB	✕ Exceeds VRAM Estimated	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce RTX 3050 6GB (Reference design)	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce RTX 3050 OC Low Profile 6G (GV-N3050OC-6GL)	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc A380 (Reference design)	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA GeForce GTX 1060	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA GeForce RTX 2060	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA GeForce RTX 3050 Discrete 6GB GDDR6	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX A2000 6GB	6 GB	✕ Exceeds VRAM	6 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P2000 (Reference design)	5 GB	✕ Exceeds VRAM	5 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P2200 (Reference design)	5 GB	✕ Exceeds VRAM	5 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
4 GB AMD Radeon R7 450 (optional)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
AMD Radeon Pro W6400	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon R9 290 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 290X (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 Fury (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 Fury X (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 Nano (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 470 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 550 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 5500 XT (Reference design), 4 GB	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 560 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 570 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 6400 Discrete 4GB GDDR6	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX 6500 XT (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon RX640	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
GeForce GTX 1050 Ti (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1630 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 1650 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 1650 SUPER (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 970 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 980 (Reference design)	4 GB	✕ Exceeds VRAM Estimated	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
Intel Arc A310 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
Intel Arc A310 LP Graphics Discrete 4GB GDDR6	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro K1200 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro K2200 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro K4200 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro M2000 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P1000 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA RTX A400	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA T1000 (Reference design)	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA T400, 4 GB	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA T600	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
RD-RX6400-E4GB/LP	4 GB	✕ Exceeds VRAM	4 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon R9 280 (Reference design)	3 GB	✕ Exceeds VRAM Estimated	3 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 280X (Reference design)	3 GB	✕ Exceeds VRAM Estimated	3 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 780 (Reference design)	3 GB	✕ Exceeds VRAM	3 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 780 Ti (Reference design)	3 GB	✕ Exceeds VRAM	3 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
2 GB AMD Radeon R5 430 (optional)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
AMD Radeon 520	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
AMD Radeon 550	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
AMD Radeon R7 240 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R7 250 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R7 260X (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 270 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 270X (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon R9 380 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
AMD Radeon RX 460 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GT 1010 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GT 1030 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GT 630 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
GeForce GT 640 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GT 710 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GT 730 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 1050 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 660 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 660 Ti (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 670 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 680 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 750 Ti (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 760 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 770 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 950 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
GeForce GTX 960 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
Intel Iris Xe MAX 100	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA GeForce GT 730	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA Quadro K620 (Reference design)	2 GB	✕ Exceeds VRAM Estimated	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
NVIDIA Quadro P400 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P600 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA Quadro P620 (Reference design)	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
NVIDIA T400, 2 GB	2 GB	✕ Exceeds VRAM	2 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
AMD Radeon 540	1 GB	✕ Exceeds VRAM Estimated	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
GeForce GT 610 (Reference design)	1 GB	✕ Exceeds VRAM Estimated	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
GeForce GT 620 (Reference design)	1 GB	✕ Exceeds VRAM Estimated	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from material whose source the record does not identify, so grade E is shown alongside B.	BE
GeForce GTX 650 (Reference design)	1 GB	✕ Exceeds VRAM	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 650 Ti (Reference design)	1 GB	✕ Exceeds VRAM	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B
GeForce GTX 750 (Reference design)	1 GB	✕ Exceeds VRAM Estimated	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory. This outcome is no stronger than the figures it was calculated from: this card's memory size is recorded here from third-party or community material, so grade D is shown alongside B.	BD
NVIDIA Quadro K600 (Reference design)	1 GB	✕ Exceeds VRAM	1 GB of memory is smaller than the 18.02 GB model file itself, so shortening the context does not make it fit in memory.	B

What each outcome means.
Outcome	Meaning
✓ Fits in VRAM	The whole model fits in this card's memory at the assumed working context length.
△ Fits at short context only	It fits at the short context length but not at the working context length.
✕ Exceeds VRAM	It does not fit in this card's memory even at the short context length. That is not the same as 'will not run': inference software can keep part of the model in system memory.
? Unverified	A value needed for the calculation is missing, so no verdict is given.

Desktops that take those cards

Only cards whose memory holds this model are listed, and only machines whose published specifications accept the card.

NVIDIA RTX A6000 — ✓ Fits in VRAM Dell Precision 3660 Tower (△ Fits with conditions)
AMD Radeon Pro W6800 — ✓ Fits in VRAM Dell Precision 3660 Tower (✓ Fits)
NVIDIA RTX 5000 Ada Generation — ✓ Fits in VRAM HP Z2 Tower G9 Workstation (△ Fits with conditions)
NVIDIA Quadro M6000 24GB (Reference design) — ✓ Fits in VRAM Estimated Dell OptiPlex 7000 Tower (2022) (△ Fits with conditions) · Dell OptiPlex 7080 Tower (2020) (△ Fits with conditions) · Dell OptiPlex 7090 Tower (2021) (△ Fits with conditions) · Dell OptiPlex Tower Plus 7010 (2023) (△ Fits with conditions) · Dell Precision 3630 Tower (△ Fits with conditions) · Dell Precision 3680 Tower (△ Fits with conditions) · HP Z2 Tower G5 Workstation (△ Fits with conditions) · HP Z2 Tower G8 Workstation (△ Fits with conditions) · HP Z2 Tower G9 Workstation (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 3 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 4 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 5 (△ Fits with conditions)
NVIDIA Quadro P6000 (Reference design) — ✓ Fits in VRAM Estimated Dell OptiPlex 7000 Tower (2022) (△ Fits with conditions) · Dell OptiPlex 7080 Tower (2020) (△ Fits with conditions) · Dell OptiPlex 7090 Tower (2021) (△ Fits with conditions) · Dell OptiPlex Tower Plus 7010 (2023) (△ Fits with conditions) · Dell Precision 3630 Tower (△ Fits with conditions) · Dell Precision 3680 Tower (△ Fits with conditions) · HP Z2 Tower G5 Workstation (△ Fits with conditions) · HP Z2 Tower G8 Workstation (△ Fits with conditions) · HP Z2 Tower G9 Workstation (△ Fits with conditions) · Lenovo ThinkCentre M90t (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 3 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 4 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 5 (△ Fits with conditions)
NVIDIA RTX 3090 — ✓ Fits in VRAM Dell Precision 3660 Tower (✓ Fits)
NVIDIA RTX 3090TI — ✓ Fits in VRAM Dell Precision 3660 Tower (✓ Fits)
NVIDIA RTX A5000 — ✓ Fits in VRAM Dell OptiPlex 7000 Tower (2022) (△ Fits with conditions) · Dell OptiPlex 7080 Tower (2020) (△ Fits with conditions) · Dell OptiPlex 7090 Tower (2021) (△ Fits with conditions) · Dell OptiPlex Tower Plus 7010 (2023) (△ Fits with conditions) · Dell Precision 3630 Tower (△ Fits with conditions) · Dell Precision 3660 Tower (△ Fits with conditions) · Dell Precision 3680 Tower (△ Fits with conditions) · HP Z2 Tower G5 Workstation (△ Fits with conditions) · HP Z2 Tower G8 Workstation (△ Fits with conditions) · HP Z2 Tower G9 Workstation (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 3 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 4 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 5 (△ Fits with conditions)
NVIDIA RTX A5500 — ✓ Fits in VRAM Dell OptiPlex 7000 Tower (2022) (△ Fits with conditions) · Dell OptiPlex 7080 Tower (2020) (△ Fits with conditions) · Dell OptiPlex 7090 Tower (2021) (△ Fits with conditions) · Dell OptiPlex Tower Plus 7010 (2023) (△ Fits with conditions) · Dell Precision 3630 Tower (△ Fits with conditions) · Dell Precision 3660 Tower (△ Fits with conditions) · Dell Precision 3680 Tower (△ Fits with conditions) · HP Z2 Tower G5 Workstation (△ Fits with conditions) · HP Z2 Tower G8 Workstation (△ Fits with conditions) · HP Z2 Tower G9 Workstation (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 3 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 4 (△ Fits with conditions) · Lenovo ThinkCentre M90t Gen 5 (△ Fits with conditions)

Verify with the manufacturer before purchasing.

Sources

Every document this page draws on, with the date it was read.
Document	Publisher	Kind	Read on
LG AI Research official EXAONE-4.0-32B-GGUF distribution repository (file listing and actual file sizes)	Hugging Face (model distribution)	primary	2026-09-24
LG AI Research official EXAONE-4.0-32B config.json (num_hidden_layers=64, num_key_value_heads=8, head_dim=128, sliding_window=4096, sliding_window_pattern="LLLG" — three local layers to one global layer, repeating; layer_types breaks down as 48 sliding-attention layers and 16 full-attention layers)	Hugging Face (model distribution)	primary	2026-09-24
LG AI Research official EXAONE 4.0 32B model card	Hugging Face (model distribution)	primary	—

We have not physically tested this combination. Data is compiled from official specification sheets and cited sources. Verify with the manufacturer before purchasing.