RTX 5090

Nvidia · 32 GB VRAM

55recorded runs

4546.8best tok/s

23sourced

How it compares — 32 GB tier

Radeon AI Pro R9700969.4 RTX 5090446.7 Intel Arc Pro B70283.1 Tesla V1004.3

Best verified tok/s per GPU in this VRAM tier — compare any two like-for-like.

Evidence: 23 sourced runs · 32 awaiting review · 1 independent source · first recorded Jun 14, 2026 · last recorded Sep 4, 2026. What these labels mean.

What fits on 32 GB — measured

Model	Quant	Peak VRAM	Headroom
Qwen3.6-27B-MTP	Q5_K_M	31.3 GB	0.7 GB
Qwen3.8-27B	FP8	30.9 GB	1.1 GB
Qwen3.8-27B	UD-Q6_K	30.3 GB	1.7 GB
Qwen3.6-27B	NVFP4	29.9 GB	2.1 GB
Qwen3.8-27B-5090-goldilocks	Q6_K	29.8 GB	2.2 GB
Qwen3.8-27B	NVFP4	29.8 GB	2.2 GB
Qwen3.6-35B-A3B-MTP	Unsloth-Dynamic-IQ4_XS	28.6 GB	3.4 GB
LFM2-350M	NVFP4	28.4 GB	3.6 GB

Peak process memory recorded by the submitter on this card — evidence of what fit, not a guarantee. Plan with the VRAM calculator.

Price history

new$9,000▲ 80.0%3 snapshots

Weekly street-price snapshots — all snapshots.

All benchmarks Compare RTX 5090 →

GPU	Model	Quant	Backend	Ctx	tok/s	prompt tok/s	VRAM GB	Source
RTX 5090 · 32 GB	NVIDIA-Nemotron-3.5-Lightning-30B-A3B	NVFP4	vllm 0.27.1	65536	446.7 ✓	—	—	src
RTX 5090 · 32 GB	Qwen3.8-27B	IQ4_XS	lucebox 298031aa4222ec61c971ed834ec8f8829ce3…	131072	435.1 ✓	1030.5	24.5	src
RTX 5090 · 32 GB	Ornith-1.0-35B-PrismaAURA-4.75bit-vllm-MTP	NVFP4	vllm v0.25.1	32768	283.1 ✓	—	—	src
RTX 5090 · 32 GB	Kimi-Linear-48B-A3B-Instruct	IQ4_XS	llama.cpp 10085 (b4aa7dd47)	32768	280.8 ✓	3733.3	25.8	src
RTX 5090 · 32 GB	Ornith-1.5-35B-A3B	NVFP4	vllm v0.27.1	2048	258.8 ✓	—	—	src
RTX 5090 · 32 GB	North-Mini-Code-1.0	UD-Q6_K_XL	llama.cpp b741 (1fd6dfe)	8192	258.4 ✓	4272.4	26.9	src
RTX 5090 · 32 GB	Qwen3.6-35B-A3B-MTP	Unsloth-Dynamic-IQ4_XS	llama.cpp 10156	65536	222.9 ✓	54.4	28.6	src
RTX 5090 · 32 GB	Laguna-XS-2.1	NVFP4	vllm v0.25.1	2048	216.7 ✓	2415.3	—	src
RTX 5090 · 32 GB	Qwen3.8-27B	NVFP4	tensorrt-llm NInfer 6e2786c5e5a58c2a8804595a9042f…	16384	213.0 ✓	9125.8	23.2	src
RTX 5090 · 32 GB	Laguna-XS-2.1	INT4	vllm v0.25.1	2048	206.2 ✓	2122.0	—	src
RTX 5090 · 32 GB	KAT-Coder-V2.5-Dev-AWQ-W4A16-ASYM	W4A16	vllm v0.27.1	2048	201.2 ✓	26.9	—	src
RTX 5090 · 32 GB	Ornith-1.0-35B	AWQ	vllm v0.25.1	65536	196.6 ✓	—	—	src
RTX 5090 · 32 GB	Qwen3.8-27B	W4A16	vllm cu129-nightly-e9d1398d9 (native DFla…	2048	137.9 ✓	—	—	src
RTX 5090 · 32 GB	Muse-Glimmer-30B	NVFP4	vllm 0.1.dev1+g075d645af	4096	124.8 ✓	2986.4	—	src
RTX 5090 · 32 GB	Qwen3.6-27B	NVFP4	vllm 0.25.1	131072	120.9 ✓	—	29.9	src
RTX 5090 · 32 GB	Huihui-Qwen3.8-27B-abliterated	NVFP4	vllm 0.27.2rc1.dev122+g8efa13b70	172800	117.7 ✓	7466.2	—	src
RTX 5090 · 32 GB	Qwen3.8-27B	W4A16	vllm v0.27.1	2048	108.0 ✓	—	—	src
RTX 5090 · 32 GB	Qwen3.8-27B-5090-goldilocks	Q6_K	llama.cpp 10437, commit 16d222fc5	131000	93.5 ✓	2038.8	29.8	src
RTX 5090 · 32 GB	Qwen3.8-27B	UD-Q6_K	llama.cpp b10478-adcf720f3	251438	79.4 ✓	541.0	30.3	src
RTX 5090 · 32 GB	Qwen3.6-27B-MTP	Q5_K_M	llama.cpp 10085 (b4aa7dd47)	131072	28.3 ✓	2581.2	31.3	src
RTX 5090 · 32 GB	DeepSeek-V4-Flash-0731	MXFP4	llama.cpp 4821 (c46ffaa5)	196608	26.4 ✓	174.5	28.3	src
RTX 5090 · 32 GB	Qwen3.8-27B	FP8	vllm 0.27.1	8192	16.8 ✓	—	30.9	src
RTX 5090 · 32 GB	Qwen3.6-27B	UD-Q5_K_XL	llama.cpp from environment.txt	131072	4.1 ✓	—	27.5	src
RTX 5090 · 32 GB	Qwen3.8-27B	NVFP4	sglang —	2048	4546.8 …	—	29.8	src
RTX 5090 · 32 GB	LFM2-350M	NVFP4	sglang —	704	2231.4 …	—	28.4	src
RTX 5090 · 32 GB	SmolLM2-135M-Instruct	Q2_K	llama.cpp —	2048	1102.1 …	75749.0	—	src
RTX 5090 · 32 GB	LFM2-24B-A2B	Q4_K_M	lmstudio —	2048	393.2 …	1718.2	—	src
RTX 5090 · 32 GB	Qwen3-30B-A3B-Instruct-2507	Q4_K_M	lmstudio —	2048	312.9 …	1108.1	—	src
RTX 5090 · 32 GB	gpt-oss-20b	MXFP4	lmstudio —	2048	307.2 …	1059.1	—	src
RTX 5090 · 32 GB	NVIDIA-Nemotron-3-Nano-30B-A3B	Q4_K_M	lmstudio —	2048	286.8 …	1114.5	—	src
RTX 5090 · 32 GB	Nemotron-3-Nano-Omni-30B-A3B-Reasoning	Q4_K_M	lmstudio —	2048	274.4 …	1208.8	—	src
RTX 5090 · 32 GB	Qwen3-Coder-30B-A3B-Instruct	Q6_K	lmstudio —	2048	263.0 …	1217.8	—	src
RTX 5090 · 32 GB	gemma-4-26B-A4B-it-qat	Q4_0	lmstudio —	2048	254.3 …	488.7	—	src
RTX 5090 · 32 GB	Qwen3.6-35B-A3B	Q4_K_M	lmstudio —	2048	243.1 …	625.2	—	src
RTX 5090 · 32 GB	NVIDIA-Nemotron-3-Nano-4B	Q8_0	lmstudio —	2048	237.8 …	1173.5	—	src
RTX 5090 · 32 GB	gemma-4-E2B-it	Q4_K_M	lmstudio —	2048	232.6 …	634.2	—	src
RTX 5090 · 32 GB	Qwen2.5-VL-7B-Instruct	Q4_K_S	lmstudio —	2048	222.4 …	2215.4	—	src
RTX 5090 · 32 GB	Qwen3-4B-Thinking-2507	Q8_0	lmstudio —	2048	219.7 …	1053.7	—	src
RTX 5090 · 32 GB	DeepSeek-R1-Distill-Qwen-7B	Q4_K_M	lmstudio —	2048	213.3 …	803.7	—	src
RTX 5090 · 32 GB	gemma-4-26B-A4B-it	Q4_K_M	lmstudio —	2048	209.8 …	458.4	—	src
RTX 5090 · 32 GB	Qwen3.5-35B-A3B	Q4_K_M	lmstudio —	2048	209.3 …	569.6	—	src
RTX 5090 · 32 GB	GLM-4.7-Flash	Q4_K_M	lmstudio —	2048	182.5 …	699.6	—	src
RTX 5090 · 32 GB	gemma-4-E4B-it	Q8_0	lmstudio —	2048	161.5 …	577.3	—	src
RTX 5090 · 32 GB	DeepSeek-R1-0528-Qwen3-8B	Q8_0	lmstudio —	2048	141.5 …	513.7	—	src
RTX 5090 · 32 GB	Bonsai-27B	Q1_0	lmstudio —	2048	136.7 …	389.4	—	src
RTX 5090 · 32 GB	gemma-3n-E4B-it-text	Q8_0	lmstudio —	2048	135.4 …	545.5	—	src
RTX 5090 · 32 GB	GLM-4.6V-Flash	Q8_0	lmstudio —	2048	132.7 …	707.8	—	src
RTX 5090 · 32 GB	gemma-4-12B-it-qat	Q4_0	lmstudio —	2048	130.8 …	368.7	—	src
RTX 5090 · 32 GB	rnj-1-instruct	Q8_0	lmstudio —	2048	123.2 …	1673.1	—	src
RTX 5090 · 32 GB	Qwen3.8-27B-MTP	NVFP4	vllm —	172800	117.7 …	7466.2	—	src
RTX 5090 · 32 GB	Phi-4-reasoning-plus	Q8_0	lmstudio —	2048	92.3 …	7981.6	—	src
RTX 5090 · 32 GB	Ministral-3-14B-Reasoning-2512	Q8_0	lmstudio —	2048	89.4 …	3183.6	—	src
RTX 5090 · 32 GB	gemma-4-12B-it	Q8_0	lmstudio —	2048	88.3 …	334.8	—	src
RTX 5090 · 32 GB	gemma-4-12B-coder-fable5-composer2.5-v1	Q8_0	lmstudio —	2048	85.3 …	330.7	—	src
RTX 5090 · 32 GB	DeepSeek-V4-Flash-0731	UD-IQ1_S	llama.cpp —	2048	18.0 …	107.1	—	src

Related reading: GPU buying guide · used GPUs · Value index · Street prices