What are the best models to run on NVIDIA RTX PRO 4000 Blackwell (24GB)?

Based on model rankings and 24 GB usable VRAM: Top Overall: Qwen 3.8 27B (INT4) ~34 tok/s. Top Coding: Qwen3.5-27B (INT4) ~34 tok/s. Optimal Daily Driver: Phi-4 Reasoning Plus (INT4) ~44 tok/s. Maximum Model Size: Falcon-40B (INT4) ~23 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 PRO 4000 Blackwell (24GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#33	Qwen 3.8 27B	27B	INT4	34 tok/s	~1.9s
#48	Agnes 3.0 Flash	33B	INT4	28 tok/s	~2.3s
#72	K2 Horizon MoVA 36B A4B	36B (4B active)	INT4	119 tok/s	~400 ms
#83	Qwen3.5-27B	27B	INT4	34 tok/s	~1.9s
#92	Sarvam-30B	32B (2.4B active)	INT4	157 tok/s	~259 ms
#94	Qwen3.6 35B A3B	35B (3B active)	INT4	136 tok/s	~327 ms
#96	Muse Glimmer 30B	30B	INT4	31 tok/s	~2.1s
#100	Phi-4 Reasoning Plus	14B	INT4 Q8	44 tok/s 30 tok/s	~987 ms ~998 ms
#101	MiMo V2 Flash	15B (309B active)	INT4 Q8	3 tok/s 2 tok/s	~20.3s ~22.3s
#108	Gemma 4 26B A4B	25.2B (3.8B active)	INT4	133 tok/s	~338 ms
#109	Gemma 4 12B	11.95B	INT4 Q8	50 tok/s 34 tok/s	~846 ms ~855 ms
#113	Qwen3.5-9B	9B	INT4 Q8 FP16	84 tok/s 53 tok/s 27 tok/s	~638 ms ~645 ms ~716 ms
#114	Qwen3.5-35B-A3B	35B (3B active)	INT4	136 tok/s	~327 ms
#118	K2 Horizon 7B	7B	INT4 Q8 FP16	101 tok/s 65 tok/s 36 tok/s	~501 ms ~506 ms ~519 ms
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	29 tok/s 17 tok/s 9 tok/s	~2.0s ~2.0s ~2.1s
#129	Gemma 4 31B	30.7B	INT4	30 tok/s	~2.1s
#131	Nemotron 3.5 Lightning	30B (3B active)	INT4	58 tok/s	~330 ms
#132	Qwen3.5-4B	4B	INT4 Q8 FP16	145 tok/s 99 tok/s 59 tok/s	~296 ms ~299 ms ~306 ms
#133	Qwen3-30B-A3B	30B (3B active)	INT4	145 tok/s	~296 ms
#136	GPT-OSS 20B	21B (3.6B active)	INT4	68 tok/s	~324 ms
#146	Granite 4.2 8B	8B	INT4 Q8 FP16	92 tok/s 58 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#150	Gemma 4 E4B	8B	INT4 Q8 FP16	92 tok/s 58 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#151	Nemotron 3.5 Lightning 30B A3B	30B (3B active)	INT4	145 tok/s	~296 ms
#153	Qwen3-32B	32B	INT4	29 tok/s	~2.2s
#156	GLM-4	32B	INT4	20 tok/s	~2.6s
#158	GLM-4V	9B	INT4 Q8 FP16	62 tok/s 44 tok/s 25 tok/s	~642 ms ~649 ms ~721 ms
#159	Gemma 3 27B	27B	INT4	34 tok/s	~1.9s
#160	Phi-4	14B	INT4 Q8	59 tok/s 36 tok/s	~981 ms ~993 ms
#162	DeepSeek-R1 32B	32B	INT4	30 tok/s	~2.2s
#164	Qwen2.5-32B	32B	INT4	29 tok/s	~2.2s
#168	Magistral Small	24B	INT4	29 tok/s	~1.7s
#169	OLMo 3 7B Think	7B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~504 ms ~510 ms ~522 ms
#170	OLMo 3.1 32B Instruct	32B	INT4	20 tok/s	~2.6s
#172	DeepSeek-R1 14B	14B	INT4 Q8	61 tok/s 36 tok/s	~981 ms ~992 ms
#173	Ministral 3 14B	14B	INT4 Q8	44 tok/s 30 tok/s	~987 ms ~998 ms
#177	Gemma 3 12B	12B	INT4 Q8 FP16	67 tok/s 41 tok/s 20 tok/s	~844 ms ~854 ms ~1.0s
#180	OLMo 3 32B Base	32B	INT4	20 tok/s	~2.6s
#181	Mistral-Small-2501	24B	INT4	37 tok/s	~1.7s
#182	Ministral 3 8B	8B	INT4 Q8 FP16	68 tok/s 48 tok/s 27 tok/s	~573 ms ~579 ms ~642 ms
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	203 tok/s 154 tok/s 100 tok/s	~160 ms ~162 ms ~165 ms
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	203 tok/s 154 tok/s 100 tok/s	~160 ms ~162 ms ~165 ms
#185	Qwen2.5-14B	14B	INT4 Q8	59 tok/s 36 tok/s	~981 ms ~993 ms
#186	Gemma 2 27B	27B	INT4	34 tok/s	~1.9s
#187	OLMo 3.1 32B Think	32B	INT4	20 tok/s	~2.6s
#188	Gemma 4 E2B	5.1B	INT4 Q8 FP16	125 tok/s 83 tok/s 48 tok/s	~371 ms ~375 ms ~384 ms
#189	Qwen3-4B	4B	INT4 Q8 FP16	145 tok/s 99 tok/s 59 tok/s	~296 ms ~299 ms ~306 ms
#190	DeepSeek-R1 8B	8B	INT4 Q8 FP16	94 tok/s 59 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#193	Ministral 3 3B	3B	INT4 Q8 FP16	120 tok/s 95 tok/s 63 tok/s	~231 ms ~233 ms ~238 ms
#195	Qwen3-14B	14B	INT4 Q8	59 tok/s 36 tok/s	~981 ms ~993 ms
#197	Gemma 2 9B	9B	INT4 Q8 FP16	84 tok/s 53 tok/s 27 tok/s	~638 ms ~645 ms ~716 ms
#198	Qwen2.5-7B	7B	INT4 Q8 FP16	101 tok/s 65 tok/s 36 tok/s	~501 ms ~506 ms ~519 ms
#199	Qwen3-8B	8B	INT4 Q8 FP16	92 tok/s 58 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#200	Gemma 3 4B	4B	INT4 Q8 FP16	145 tok/s 99 tok/s 59 tok/s	~296 ms ~299 ms ~306 ms
#201	DeepSeek-R1 7B	7B	INT4 Q8 FP16	104 tok/s 66 tok/s 36 tok/s	~500 ms ~506 ms ~518 ms
#202	Llama 3.1 8B	8B	INT4 Q8 FP16	92 tok/s 58 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	217 tok/s 168 tok/s 112 tok/s	~138 ms ~139 ms ~142 ms
#205	Command R	35B	INT4	19 tok/s	~2.8s
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	311 tok/s 293 tok/s 257 tok/s	~28 ms ~28 ms ~29 ms
#208	Llama 3 8B	8B	INT4 Q8 FP16	92 tok/s 58 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	298 tok/s 261 tok/s 205 tok/s	~53 ms ~54 ms ~55 ms
#210	Phi-4-Mini	3.8B	INT4 Q8 FP16	149 tok/s 103 tok/s 61 tok/s	~283 ms ~286 ms ~292 ms
#212	Ministral-8B-2410	8B	INT4 Q8 FP16	92 tok/s 58 tok/s 32 tok/s	~569 ms ~575 ms ~589 ms
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	228 tok/s 180 tok/s 122 tok/s	~123 ms ~124 ms ~127 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	169 tok/s 121 tok/s 74 tok/s	~228 ms ~230 ms ~236 ms
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	278 tok/s 237 tok/s 178 tok/s	~69 ms ~70 ms ~71 ms
#216	Phi-3-mini	3.8B	INT4 Q8 FP16	149 tok/s 103 tok/s 61 tok/s	~283 ms ~286 ms ~292 ms
#217	Gemma 3n E2B IT	6B (2B active)	INT4 Q8 FP16	119 tok/s 103 tok/s 76 tok/s	~174 ms ~175 ms ~179 ms
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8 FP16	98 tok/s 63 tok/s 35 tok/s	~521 ms ~527 ms ~539 ms
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	260 tok/s 216 tok/s 157 tok/s	~86 ms ~87 ms ~89 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	203 tok/s 154 tok/s 100 tok/s	~160 ms ~162 ms ~165 ms
#221	Phi-3-medium	14B	INT4 Q8	59 tok/s 36 tok/s	~981 ms ~993 ms
#222	OLMo 3 7B Instruct	7B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~504 ms ~510 ms ~522 ms
#223	Yi-34B	34B	INT4	19 tok/s	~2.7s
#224	Qwen2-7B	7B	INT4 Q8 FP16	101 tok/s 65 tok/s 36 tok/s	~501 ms ~506 ms ~519 ms
#225	Phi-3-small	7B	INT4 Q8 FP16	101 tok/s 65 tok/s 36 tok/s	~501 ms ~506 ms ~519 ms
#226	Gemma 3 1B	1B	INT4 Q8 FP16	260 tok/s 216 tok/s 157 tok/s	~86 ms ~87 ms ~89 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8 FP16	98 tok/s 63 tok/s 35 tok/s	~521 ms ~527 ms ~539 ms
#229	Gemma 1 7B	7B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~504 ms ~510 ms ~522 ms
#230	Gemma 1 2B	2B	INT4 Q8 FP16	214 tok/s 159 tok/s 102 tok/s	~160 ms ~162 ms ~165 ms
#231	ChatGLM3-6B	6B	INT4 Q8 FP16	82 tok/s 60 tok/s 37 tok/s	~436 ms ~440 ms ~451 ms
#232	ChatGLM2-6B	6B	INT4 Q8 FP16	82 tok/s 60 tok/s 37 tok/s	~436 ms ~440 ms ~451 ms
#233	ChatGLM-6B	6B	INT4 Q8 FP16	82 tok/s 60 tok/s 37 tok/s	~436 ms ~440 ms ~451 ms
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	308 tok/s 275 tok/s 221 tok/s	~45 ms ~46 ms ~46 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	308 tok/s 275 tok/s 221 tok/s	~45 ms ~46 ms ~46 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	304 tok/s 285 tok/s 248 tok/s	~30 ms ~30 ms ~31 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	304 tok/s 285 tok/s 248 tok/s	~30 ms ~30 ms ~31 ms
-	Falcon-1B	1B	INT4 Q8 FP16	270 tok/s 223 tok/s 160 tok/s	~86 ms ~87 ms ~89 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	260 tok/s 216 tok/s 157 tok/s	~86 ms ~87 ms ~89 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	228 tok/s 180 tok/s 122 tok/s	~123 ms ~124 ms ~127 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	228 tok/s 180 tok/s 122 tok/s	~123 ms ~124 ms ~127 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	217 tok/s 168 tok/s 112 tok/s	~138 ms ~139 ms ~142 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	213 tok/s 164 tok/s 108 tok/s	~145 ms ~147 ms ~150 ms
-	Falcon-3B	3B	INT4 Q8 FP16	177 tok/s 125 tok/s 75 tok/s	~228 ms ~230 ms ~235 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	175 tok/s 151 tok/s 112 tok/s	~110 ms ~110 ms ~113 ms
-	Phi-1	1.3B	INT4 Q8 FP16	175 tok/s 151 tok/s 112 tok/s	~110 ms ~110 ms ~113 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	175 tok/s 151 tok/s 112 tok/s	~110 ms ~110 ms ~113 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	171 tok/s 122 tok/s 74 tok/s	~228 ms ~230 ms ~236 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	169 tok/s 121 tok/s 74 tok/s	~228 ms ~230 ms ~236 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	169 tok/s 121 tok/s 74 tok/s	~228 ms ~230 ms ~236 ms
-	ERNIE-4.5-21B-A3B	21B (3B active)	INT4 Q8	151 tok/s 101 tok/s	~277 ms ~327 ms
-	ERNIE-4.5-21B-A3B-Base	21B (3B active)	INT4 Q8	151 tok/s 101 tok/s	~277 ms ~327 ms
-	ERNIE-4.5-VL-28B-A3B	28B (3B active)	INT4	146 tok/s	~291 ms
-	ERNIE-4.5-VL-28B-A3B-Base	28B (3B active)	INT4	146 tok/s	~291 ms
-	Hy-MT2 30B A3B	30B (3B active)	INT4	145 tok/s	~296 ms
-	Spark X2.5 4B	4B	INT4 Q8 FP16	145 tok/s 99 tok/s 59 tok/s	~296 ms ~299 ms ~306 ms
-	Phi-2	2.7B	INT4 Q8 FP16	125 tok/s 101 tok/s 68 tok/s	~210 ms ~212 ms ~217 ms
-	MaLLaM-3B	3B	INT4 Q8 FP16	120 tok/s 95 tok/s 63 tok/s	~231 ms ~233 ms ~238 ms
-	SmolLM3 3B	3B	INT4 Q8 FP16	120 tok/s 95 tok/s 63 tok/s	~231 ms ~233 ms ~238 ms
-	Falcon-7B	7B	INT4 Q8 FP16	105 tok/s 66 tok/s 37 tok/s	~500 ms ~506 ms ~518 ms
-	Falcon3-7B	7B	INT4 Q8 FP16	101 tok/s 65 tok/s 36 tok/s	~501 ms ~506 ms ~519 ms
-	Hy-MT2 7B	7B	INT4 Q8 FP16	101 tok/s 65 tok/s 36 tok/s	~501 ms ~506 ms ~519 ms
-	Mistral-7B-v0.1	7.3B	INT4 Q8 FP16	98 tok/s 63 tok/s 35 tok/s	~521 ms ~527 ms ~539 ms
-	LensVLM 9B	9B	INT4 Q8 FP16	84 tok/s 53 tok/s 27 tok/s	~638 ms ~645 ms ~716 ms
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8 FP16	84 tok/s 53 tok/s 27 tok/s	~638 ms ~645 ms ~716 ms
-	ChatGLM3-6B-32K	6B	INT4 Q8 FP16	82 tok/s 60 tok/s 37 tok/s	~436 ms ~440 ms ~451 ms
-	Yi-6B	6B	INT4 Q8 FP16	82 tok/s 60 tok/s 37 tok/s	~436 ms ~440 ms ~451 ms
-	Falcon3-10B	10B	INT4 Q8 FP16	78 tok/s 48 tok/s 25 tok/s	~706 ms ~714 ms ~793 ms
-	Kimi-VL-A3B-Instruct	16B (3B active)	INT4 Q8	78 tok/s 65 tok/s	~273 ms ~297 ms
-	Falcon2-11B	11B	INT4 Q8 FP16	75 tok/s 45 tok/s 22 tok/s	~774 ms ~783 ms ~939 ms
-	Hunyuan Lite	7B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~504 ms ~510 ms ~522 ms
-	OLMo 3 7B Base	7B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~504 ms ~510 ms ~522 ms
-	SEA-LION-7B	7.1B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~511 ms ~516 ms ~529 ms
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8 FP16	74 tok/s 53 tok/s 32 tok/s	~511 ms ~516 ms ~529 ms
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8 FP16	68 tok/s 48 tok/s 27 tok/s	~573 ms ~579 ms ~642 ms
-	Typhoon-2-8B	8B	INT4 Q8 FP16	68 tok/s 48 tok/s 27 tok/s	~573 ms ~579 ms ~642 ms
-	GLM-4-9B	9B	INT4 Q8 FP16	62 tok/s 44 tok/s 25 tok/s	~642 ms ~649 ms ~721 ms
-	GLM-4-9B-Chat	9B	INT4 Q8 FP16	62 tok/s 44 tok/s 25 tok/s	~642 ms ~649 ms ~721 ms
-	GLM-4-9B-Chat-1M	9B	INT4 Q8 FP16	62 tok/s 44 tok/s 25 tok/s	~642 ms ~649 ms ~721 ms
-	Yi-9B	9B	INT4 Q8 FP16	62 tok/s 44 tok/s 25 tok/s	~642 ms ~649 ms ~721 ms
-	Sahabat-AI-Gemma2-9B	9.2B	INT4 Q8 FP16	61 tok/s 43 tok/s 24 tok/s	~656 ms ~663 ms ~736 ms
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4 Q8 FP16	61 tok/s 43 tok/s 24 tok/s	~656 ms ~663 ms ~736 ms
-	GreenMind-14B-R1	14B	INT4 Q8	44 tok/s 30 tok/s	~987 ms ~998 ms
-	Falcon-40B	40B	INT4	23 tok/s	~3.0s

NVIDIA RTX PRO 4000 Blackwell (24GB)

VRAM Scaling by Context Length

Best Models to Run on NVIDIA RTX PRO 4000 Blackwell (24GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions