NVIDIA · Ada

What runs on a RTX 4090

24 GB at 1008 GB/s. The largest model that fits at Q4_K_M is Qwen3 32B, at roughly 30.5 tokens per second.

Model

Hardware

Quantisation

Context 8K

KV cache precision

Runtime

Batch

GPUs

Monitor plugged into this GPU

Runs comfortably

6.4 GB of 21.8 GB · 30%

022 GB

Weights 4.6 GB

KV cache 1.0 GB

Runtime overhead 0.8 GB

Llama 3.1 8B at Q4_K_M leaves 15.3 GB spare on a RTX 4090. There is room to raise the context length or move up a quantisation level.

Generation116tok/s

Prompt processing4315tok/s

Max context128Ktokens

KV per 1K tokens0GB

Every quantisation of Llama 3.1 8B on a RTX 4090

Highlighted row is the highest quality that still fits at 8K context.

Quantisation	bpw	Weights	Total	Verdict	Max ctx	tok/s
FP16 / BF16	16.00	15.0 GB	16.8 GB	Runs comfortably	48K	38.8
INT8 / W8A8	8.50	7.7 GB	9.5 GB	Runs comfortably	106K	72.6
Q8_0 (GGUF)	8.50	7.7 GB	9.5 GB	Runs comfortably	106K	72.6
FP8 (E4M3)	8.00	7.5 GB	9.3 GB	Runs comfortably	108K	74.7
Q6_K	6.56	6.1 GB	8.0 GB	Runs comfortably	118K	89.6
AWQ 4-bit	4.25	5.4 GB	7.2 GB	Runs comfortably	124K	100
GPTQ 4-bit	4.25	5.4 GB	7.2 GB	Runs comfortably	124K	100
MXFP4	4.25	5.4 GB	7.2 GB	Runs comfortably	124K	100
Q5_K_M	5.67	5.3 GB	7.1 GB	Runs comfortably	125K	102
Q5_K_S	5.52	5.2 GB	7.0 GB	Runs comfortably	126K	105
Q4_K_M	4.85	4.6 GB	6.4 GB	Runs comfortably	128K	116
Q4_K_S	4.58	4.4 GB	6.2 GB	Runs comfortably	128K	121
Q4_0	4.55	4.4 GB	6.2 GB	Runs comfortably	128K	121
IQ4_XS	4.25	4.1 GB	5.9 GB	Runs comfortably	128K	127
Q3_K_M	3.91	3.8 GB	5.7 GB	Runs comfortably	128K	135
IQ3_M	3.70	3.7 GB	5.5 GB	Runs comfortably	128K	141
IQ3_XXS	3.06	3.2 GB	5.0 GB	Runs comfortably	128K	160
Q2_K	2.63	2.8 GB	4.6 GB	Runs comfortably	128K	176
IQ2_XXS	2.06	2.3 GB	4.2 GB	Runs comfortably	128K	204
IQ1_M	1.75	2.1 GB	3.9 GB	Runs comfortably	128K	223

Q4_K_M at 8K context. Bandwidth sets the speed; capacity sets the ceiling.

Model	Params	Weights	Verdict	Max ctx	tok/s
DeepSeek-R1 671B-A37B	671B	379.0 GB	Won't fit	—	1.88
Qwen3 235B-A22B	235.1B	132.8 GB	Won't fit	—	3.23
gpt-oss 120B-A5.1B	116.8B	66.0 GB	Won't fit	—	16.4
Llama 4 Scout 109B-A17B	109B	61.7 GB	Won't fit	—	5.18
Qwen2.5 72B	72.7B	41.2 GB	Won't fit	—	1.53
Llama 3.1 70B	70.5B	40.0 GB	Won't fit	—	1.60
Mixtral 8x7B	46.7B	26.4 GB	Won't fit	—	15.9
Command R 35B	35.0B	20.1 GB	Won't fit	640	3.00
Yi-1.5 34B	34.4B	19.5 GB	Won't fit	6K	19.7
Qwen3 32B	32.8B	18.6 GB	Fits, but tight	9K	30.5
Qwen2.5-Coder 32B	32.8B	18.6 GB	Fits, but tight	9K	30.5
DeepSeek-R1-Distill-Qwen 32B	32.8B	18.6 GB	Fits, but tight	9K	30.5
Qwen3 30B-A3B	30.5B	17.3 GB	Runs comfortably	39K	118
Gemma 3 27B	27.4B	15.6 GB	Runs comfortably	57K	36.5
Mistral Small 24B	23.6B	13.4 GB	Runs comfortably	32K	42.6
gpt-oss 20B-A3.6B	20.9B	11.9 GB	Runs comfortably	128K	188
Qwen3 14B	14.8B	8.5 GB	Runs comfortably	80K	65.5
Phi-4 14B	14.7B	8.4 GB	Runs comfortably	16K	65.0
Mistral NeMo 12B	12.3B	7.0 GB	Runs comfortably	89K	77.6
Gemma 3 12B	12.2B	7.0 GB	Runs comfortably	128K	78.5
GLM-4 9B	9.4B	5.4 GB	Runs comfortably	128K	106
Qwen3 8B	8.2B	4.7 GB	Runs comfortably	115K	112
Llama 3.1 8B	8.0B	4.6 GB	Runs comfortably	128K	116
DeepSeek-R1-Distill-Llama 8B	8.0B	4.6 GB	Runs comfortably	128K	116
Qwen2.5 7B	7.6B	4.4 GB	Runs comfortably	128K	128
Mistral 7B v0.3	7.3B	4.1 GB	Runs comfortably	32K	128
Gemma 3 4B	4.3B	2.5 GB	Runs comfortably	128K	215
Qwen3 4B	4.0B	2.3 GB	Runs comfortably	128K	202
Llama 3.2 3B	3.2B	1.8 GB	Runs comfortably	128K	254
Llama 3.2 1B	1.2B	0.7 GB	Runs comfortably	128K	669