What are the best models to run on NVIDIA RTX 4500 Blackwell (32GB)?

Based on model rankings and 32 GB usable VRAM: Top Overall: Qwen 3.8 27B (INT4) ~44 tok/s. Top Coding: Qwen3.5-27B (INT4) ~44 tok/s. Optimal Daily Driver: Phi-4 Reasoning Plus (INT4) ~56 tok/s. Maximum Model Size: Kimi Linear 48B A3B Instruct (INT4) ~60 tok/s.

How are VRAM, throughput, and latency calculated?

Calculations are determined using apxml.com modeling, combining LLM architecture analysis with GPU hardware physics. VRAM and Context Scaling: Memory requirements are determined from model architecture, including parameter count, layer configuration, and attention structure. The KV cache footprint scales deterministically based on context length, batch size, and cache precision. Throughput (TPS): Token generation is calculated in the memory-bandwidth-bound decode phase, measuring the time required to stream model weights and KV cache relative to hardware memory bandwidth. Latency (TTFT): Time to first token is calculated from the compute operations required during prompt prefill relative to hardware tensor compute capacity.

How accurate are these estimates?

Calculations are calibrated against empirical ground-truth benchmark runs across diverse architectures and hardware. Because the underlying mechanics are benchmark-calibrated, estimates remain accurate across different configurations without needing to run physical tests for every scenario. VRAM Accuracy: Memory estimates are deterministic and closely match actual allocation. Minor differences typically stay within 3% to 5% due to CUDA runtime overhead or memory allocator fragmentation. Speed and Latency Accuracy: Throughput and TTFT typically vary within 5% to 10% depending on the specific inference runtime, kernel implementations, and system power limits.

Best Models to Run on NVIDIA RTX 4500 Blackwell (32GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#33	Qwen 3.8 27B	27B	INT4 Q8	44 tok/s 24 tok/s	~1.5s ~1.6s
#48	Agnes 3.0 Flash	33B	INT4	37 tok/s	~1.8s
#72	K2 Horizon MoVA 36B A4B	36B (4B active)	INT4	150 tok/s	~293 ms
#83	Qwen3.5-27B	27B	INT4 Q8	44 tok/s 24 tok/s	~1.5s ~1.6s
#92	Sarvam-30B	32B (2.4B active)	INT4	184 tok/s	~206 ms
#94	Qwen3.6 35B A3B	35B (3B active)	INT4	169 tok/s	~239 ms
#96	Muse Glimmer 30B	30B	INT4 Q8	40 tok/s 21 tok/s	~1.6s ~1.9s
#100	Phi-4 Reasoning Plus	14B	INT4 Q8 FP16	56 tok/s 39 tok/s 20 tok/s	~781 ms ~789 ms ~944 ms
#101	MiMo V2 Flash	15B (309B active)	INT4 Q8	4 tok/s 2 tok/s	~16.0s ~16.2s
#108	Gemma 4 26B A4B	25.2B (3.8B active)	INT4 Q8	158 tok/s 112 tok/s	~268 ms ~293 ms
#109	Gemma 4 12B	11.95B	INT4 Q8 FP16	64 tok/s 44 tok/s 25 tok/s	~669 ms ~676 ms ~750 ms
#113	Qwen3.5-9B	9B	INT4 Q8 FP16	104 tok/s 67 tok/s 37 tok/s	~505 ms ~510 ms ~522 ms
#114	Qwen3.5-35B-A3B	35B (3B active)	INT4	169 tok/s	~239 ms
#118	K2 Horizon 7B	7B	INT4 Q8 FP16	124 tok/s 82 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	38 tok/s 23 tok/s 12 tok/s	~1.6s ~1.6s ~1.6s
#129	Gemma 4 31B	30.7B	INT4 Q8	39 tok/s 20 tok/s	~1.7s ~2.0s
#131	Nemotron 3.5 Lightning	30B (3B active)	INT4	78 tok/s	~242 ms
#132	Qwen3.5-4B	4B	INT4 Q8 FP16	171 tok/s 122 tok/s 74 tok/s	~235 ms ~238 ms ~243 ms
#133	Qwen3-30B-A3B	30B (3B active)	INT4 Q8	171 tok/s 119 tok/s	~235 ms ~277 ms
#136	GPT-OSS 20B	21B (3.6B active)	INT4 Q8	86 tok/s 71 tok/s	~257 ms ~280 ms
#146	Granite 4.2 8B	8B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#150	Gemma 4 E4B	8B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#151	Nemotron 3.5 Lightning 30B A3B	30B (3B active)	INT4 Q8	171 tok/s 119 tok/s	~235 ms ~277 ms
#153	Qwen3-32B	32B	INT4	38 tok/s	~1.7s
#155	Mistral Large 3	41B (675B active)	INT4	2 tok/s	~41.0s
#156	GLM-4	32B	INT4	30 tok/s	~1.7s
#158	GLM-4V	9B	INT4 Q8 FP16	78 tok/s 56 tok/s 34 tok/s	~508 ms ~513 ms ~525 ms
#159	Gemma 3 27B	27B	INT4 Q8	44 tok/s 24 tok/s	~1.5s ~1.6s
#160	Phi-4	14B	INT4 Q8 FP16	75 tok/s 46 tok/s 24 tok/s	~776 ms ~785 ms ~870 ms
#162	DeepSeek-R1 32B	32B	INT4 Q8	39 tok/s 20 tok/s	~1.7s ~2.1s
#164	Qwen2.5-32B	32B	INT4	38 tok/s	~1.7s
#168	Magistral Small	24B	INT4 Q8	37 tok/s 23 tok/s	~1.3s ~1.4s
#169	OLMo 3 7B Think	7B	INT4 Q8 FP16	93 tok/s 68 tok/s 42 tok/s	~399 ms ~403 ms ~413 ms
#170	OLMo 3.1 32B Instruct	32B	INT4	30 tok/s	~1.7s
#172	DeepSeek-R1 14B	14B	INT4 Q8 FP16	77 tok/s 47 tok/s 24 tok/s	~776 ms ~784 ms ~870 ms
#173	Ministral 3 14B	14B	INT4 Q8 FP16	56 tok/s 39 tok/s 20 tok/s	~781 ms ~789 ms ~944 ms
#177	Gemma 3 12B	12B	INT4 Q8 FP16	85 tok/s 53 tok/s 29 tok/s	~668 ms ~675 ms ~691 ms
#180	OLMo 3 32B Base	32B	INT4	30 tok/s	~1.7s
#181	Mistral-Small-2501	24B	INT4 Q8	48 tok/s 27 tok/s	~1.3s ~1.4s
#182	Ministral 3 8B	8B	INT4 Q8 FP16	85 tok/s 62 tok/s 37 tok/s	~453 ms ~458 ms ~468 ms
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	229 tok/s 180 tok/s 122 tok/s	~128 ms ~129 ms ~132 ms
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	229 tok/s 180 tok/s 122 tok/s	~128 ms ~129 ms ~132 ms
#185	Qwen2.5-14B	14B	INT4 Q8 FP16	75 tok/s 46 tok/s 24 tok/s	~776 ms ~785 ms ~870 ms
#186	Gemma 2 27B	27B	INT4 Q8	44 tok/s 24 tok/s	~1.5s ~1.6s
#187	OLMo 3.1 32B Think	32B	INT4	30 tok/s	~1.7s
#188	Gemma 4 E2B	5.1B	INT4 Q8 FP16	150 tok/s 103 tok/s 61 tok/s	~295 ms ~298 ms ~304 ms
#189	Qwen3-4B	4B	INT4 Q8 FP16	171 tok/s 122 tok/s 74 tok/s	~235 ms ~238 ms ~243 ms
#190	DeepSeek-R1 8B	8B	INT4 Q8 FP16	116 tok/s 75 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#193	Ministral 3 3B	3B	INT4 Q8 FP16	144 tok/s 117 tok/s 80 tok/s	~183 ms ~185 ms ~189 ms
#195	Qwen3-14B	14B	INT4 Q8 FP16	75 tok/s 46 tok/s 24 tok/s	~776 ms ~785 ms ~870 ms
#196	Kimi Linear 48B A3B Instruct	48B (3B active)	INT4	60 tok/s	~306 ms
#197	Gemma 2 9B	9B	INT4 Q8 FP16	104 tok/s 67 tok/s 37 tok/s	~505 ms ~510 ms ~522 ms
#198	Qwen2.5-7B	7B	INT4 Q8 FP16	124 tok/s 82 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
#199	Qwen3-8B	8B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#200	Gemma 3 4B	4B	INT4 Q8 FP16	171 tok/s 122 tok/s 74 tok/s	~235 ms ~238 ms ~243 ms
#201	DeepSeek-R1 7B	7B	INT4 Q8 FP16	126 tok/s 83 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
#202	Llama 3.1 8B	8B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	242 tok/s 195 tok/s 136 tok/s	~111 ms ~112 ms ~114 ms
#205	Command R	35B	INT4	28 tok/s	~1.9s
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	324 tok/s 309 tok/s 279 tok/s	~24 ms ~24 ms ~24 ms
#208	Llama 3 8B	8B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	313 tok/s 282 tok/s 231 tok/s	~44 ms ~44 ms ~45 ms
#210	Phi-4-Mini	3.8B	INT4 Q8 FP16	175 tok/s 126 tok/s 77 tok/s	~225 ms ~227 ms ~232 ms
#211	Mixtral-8x7B-v0.1	46.7B (7B active)	INT4	106 tok/s	~498 ms
#212	Ministral-8B-2410	8B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~450 ms ~455 ms ~466 ms
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	253 tok/s 206 tok/s 147 tok/s	~99 ms ~100 ms ~102 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	196 tok/s 145 tok/s 92 tok/s	~182 ms ~183 ms ~187 ms
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	297 tok/s 260 tok/s 204 tok/s	~56 ms ~57 ms ~58 ms
#216	Phi-3-mini	3.8B	INT4 Q8 FP16	175 tok/s 126 tok/s 77 tok/s	~225 ms ~227 ms ~232 ms
#217	Gemma 3n E2B IT	6B (2B active)	INT4 Q8 FP16	144 tok/s 126 tok/s 95 tok/s	~139 ms ~140 ms ~142 ms
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8 FP16	121 tok/s 79 tok/s 45 tok/s	~413 ms ~417 ms ~427 ms
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	281 tok/s 241 tok/s 183 tok/s	~70 ms ~70 ms ~72 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	229 tok/s 180 tok/s 122 tok/s	~128 ms ~129 ms ~132 ms
#221	Phi-3-medium	14B	INT4 Q8 FP16	75 tok/s 46 tok/s 24 tok/s	~776 ms ~785 ms ~870 ms
#222	OLMo 3 7B Instruct	7B	INT4 Q8 FP16	93 tok/s 68 tok/s 42 tok/s	~399 ms ~403 ms ~413 ms
#223	Yi-34B	34B	INT4	28 tok/s	~1.8s
#224	Qwen2-7B	7B	INT4 Q8 FP16	124 tok/s 82 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
#225	Phi-3-small	7B	INT4 Q8 FP16	124 tok/s 82 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
#226	Gemma 3 1B	1B	INT4 Q8 FP16	281 tok/s 241 tok/s 183 tok/s	~70 ms ~70 ms ~72 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8 FP16	121 tok/s 79 tok/s 45 tok/s	~413 ms ~417 ms ~427 ms
#229	Gemma 1 7B	7B	INT4 Q8 FP16	93 tok/s 68 tok/s 42 tok/s	~399 ms ~403 ms ~413 ms
#230	Gemma 1 2B	2B	INT4 Q8 FP16	239 tok/s 186 tok/s 125 tok/s	~128 ms ~129 ms ~132 ms
#231	ChatGLM3-6B	6B	INT4 Q8 FP16	102 tok/s 76 tok/s 47 tok/s	~345 ms ~349 ms ~357 ms
#232	ChatGLM2-6B	6B	INT4 Q8 FP16	102 tok/s 76 tok/s 47 tok/s	~345 ms ~349 ms ~357 ms
#233	ChatGLM-6B	6B	INT4 Q8 FP16	102 tok/s 76 tok/s 47 tok/s	~345 ms ~349 ms ~357 ms
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	322 tok/s 294 tok/s 246 tok/s	~37 ms ~38 ms ~38 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	322 tok/s 294 tok/s 246 tok/s	~37 ms ~38 ms ~38 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	318 tok/s 303 tok/s 270 tok/s	~25 ms ~26 ms ~26 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	318 tok/s 303 tok/s 270 tok/s	~25 ms ~26 ms ~26 ms
-	Falcon-1B	1B	INT4 Q8 FP16	290 tok/s 247 tok/s 187 tok/s	~69 ms ~70 ms ~71 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	281 tok/s 241 tok/s 183 tok/s	~70 ms ~70 ms ~72 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	253 tok/s 206 tok/s 147 tok/s	~99 ms ~100 ms ~102 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	253 tok/s 206 tok/s 147 tok/s	~99 ms ~100 ms ~102 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	242 tok/s 195 tok/s 136 tok/s	~111 ms ~112 ms ~114 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	238 tok/s 190 tok/s 131 tok/s	~116 ms ~117 ms ~120 ms
-	Falcon-3B	3B	INT4 Q8 FP16	204 tok/s 149 tok/s 94 tok/s	~181 ms ~183 ms ~187 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	201 tok/s 177 tok/s 136 tok/s	~88 ms ~89 ms ~90 ms
-	Phi-1	1.3B	INT4 Q8 FP16	201 tok/s 177 tok/s 136 tok/s	~88 ms ~89 ms ~90 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	201 tok/s 177 tok/s 136 tok/s	~88 ms ~89 ms ~90 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	198 tok/s 146 tok/s 93 tok/s	~182 ms ~183 ms ~187 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	196 tok/s 145 tok/s 92 tok/s	~182 ms ~183 ms ~187 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	196 tok/s 145 tok/s 92 tok/s	~182 ms ~183 ms ~187 ms
-	ERNIE-4.5-21B-A3B	21B (3B active)	INT4 Q8	177 tok/s 136 tok/s	~220 ms ~222 ms
-	ERNIE-4.5-21B-A3B-Base	21B (3B active)	INT4 Q8	177 tok/s 136 tok/s	~220 ms ~222 ms
-	ERNIE-4.5-VL-28B-A3B	28B (3B active)	INT4 Q8	172 tok/s 126 tok/s	~231 ms ~252 ms
-	ERNIE-4.5-VL-28B-A3B-Base	28B (3B active)	INT4 Q8	172 tok/s 126 tok/s	~231 ms ~252 ms
-	Hy-MT2 30B A3B	30B (3B active)	INT4 Q8	171 tok/s 119 tok/s	~235 ms ~277 ms
-	Spark X2.5 4B	4B	INT4 Q8 FP16	171 tok/s 122 tok/s 74 tok/s	~235 ms ~238 ms ~243 ms
-	Phi-2	2.7B	INT4 Q8 FP16	150 tok/s 123 tok/s 86 tok/s	~167 ms ~169 ms ~172 ms
-	MaLLaM-3B	3B	INT4 Q8 FP16	144 tok/s 117 tok/s 80 tok/s	~183 ms ~185 ms ~189 ms
-	SmolLM3 3B	3B	INT4 Q8 FP16	144 tok/s 117 tok/s 80 tok/s	~183 ms ~185 ms ~189 ms
-	Falcon-7B	7B	INT4 Q8 FP16	128 tok/s 84 tok/s 47 tok/s	~396 ms ~401 ms ~410 ms
-	Falcon3-7B	7B	INT4 Q8 FP16	124 tok/s 82 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
-	Hy-MT2 7B	7B	INT4 Q8 FP16	124 tok/s 82 tok/s 47 tok/s	~397 ms ~401 ms ~410 ms
-	Mistral-7B-v0.1	7.3B	INT4 Q8 FP16	121 tok/s 79 tok/s 45 tok/s	~413 ms ~417 ms ~427 ms
-	LensVLM 9B	9B	INT4 Q8 FP16	104 tok/s 67 tok/s 37 tok/s	~505 ms ~510 ms ~522 ms
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8 FP16	104 tok/s 67 tok/s 37 tok/s	~505 ms ~510 ms ~522 ms
-	ChatGLM3-6B-32K	6B	INT4 Q8 FP16	102 tok/s 76 tok/s 47 tok/s	~345 ms ~349 ms ~357 ms
-	Yi-6B	6B	INT4 Q8 FP16	102 tok/s 76 tok/s 47 tok/s	~345 ms ~349 ms ~357 ms
-	Falcon3-10B	10B	INT4 Q8 FP16	97 tok/s 61 tok/s 34 tok/s	~559 ms ~565 ms ~578 ms
-	Kimi-VL-A3B-Instruct	16B (3B active)	INT4 Q8	97 tok/s 86 tok/s	~216 ms ~218 ms
-	Falcon2-11B	11B	INT4 Q8 FP16	94 tok/s 58 tok/s 32 tok/s	~613 ms ~619 ms ~634 ms
-	Hunyuan Lite	7B	INT4 Q8 FP16	93 tok/s 68 tok/s 42 tok/s	~399 ms ~403 ms ~413 ms
-	OLMo 3 7B Base	7B	INT4 Q8 FP16	93 tok/s 68 tok/s 42 tok/s	~399 ms ~403 ms ~413 ms
-	SEA-LION-7B	7.1B	INT4 Q8 FP16	92 tok/s 67 tok/s 41 tok/s	~405 ms ~409 ms ~418 ms
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8 FP16	92 tok/s 67 tok/s 41 tok/s	~405 ms ~409 ms ~418 ms
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8 FP16	85 tok/s 62 tok/s 37 tok/s	~453 ms ~458 ms ~468 ms
-	Typhoon-2-8B	8B	INT4 Q8 FP16	85 tok/s 62 tok/s 37 tok/s	~453 ms ~458 ms ~468 ms
-	GLM-4-9B	9B	INT4 Q8 FP16	78 tok/s 56 tok/s 34 tok/s	~508 ms ~513 ms ~525 ms
-	GLM-4-9B-Chat	9B	INT4 Q8 FP16	78 tok/s 56 tok/s 34 tok/s	~508 ms ~513 ms ~525 ms
-	GLM-4-9B-Chat-1M	9B	INT4 Q8 FP16	78 tok/s 56 tok/s 34 tok/s	~508 ms ~513 ms ~525 ms
-	Yi-9B	9B	INT4 Q8 FP16	78 tok/s 56 tok/s 34 tok/s	~508 ms ~513 ms ~525 ms
-	Sahabat-AI-Gemma2-9B	9.2B	INT4 Q8 FP16	77 tok/s 55 tok/s 33 tok/s	~519 ms ~524 ms ~536 ms
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4 Q8 FP16	77 tok/s 55 tok/s 33 tok/s	~519 ms ~524 ms ~536 ms
-	GreenMind-14B-R1	14B	INT4 Q8 FP16	56 tok/s 39 tok/s 20 tok/s	~781 ms ~789 ms ~944 ms
-	Falcon-40B	40B	INT4	32 tok/s	~2.2s

NVIDIA RTX 4500 Blackwell (32GB)

VRAM Scaling by Context Length

Best Models to Run on NVIDIA RTX 4500 Blackwell (32GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions