What are the best models to run on NVIDIA RTX 3060 (12GB)?

Based on model rankings and 12 GB usable VRAM: Top Overall: Phi-4 Reasoning Plus (INT4) ~22 tok/s. Top Coding: K2 Horizon 7B (INT4) ~62 tok/s. Optimal Daily Driver: Qwen3.5-9B (INT4) ~50 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 LLMs to Run on NVIDIA RTX 3060 (12GB)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#102	Phi-4 Reasoning Plus	14B	INT4	22 tok/s	~3.6s
#121	Qwen3.5-9B	9B	INT4 Q8	50 tok/s 29 tok/s	~2.0s ~2.1s
#127	K2 Horizon 7B	7B	INT4 Q8	62 tok/s 38 tok/s	~1.5s ~1.5s
#130	Gemma 4 12B	11.95B	INT4	27 tok/s	~2.8s
#136	Qwen3.5-4B	4B	INT4 Q8 FP16	95 tok/s 61 tok/s 34 tok/s	~900 ms ~906 ms ~919 ms
#140	GLM-4V	9B	INT4	36 tok/s	~2.0s
#145	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	16 tok/s 9 tok/s 5 tok/s	~6.2s ~6.2s ~6.8s
#152	Granite 4.2 8B	8B	INT4 Q8	56 tok/s 32 tok/s	~1.7s ~1.9s
#157	Gemma 4 E4B	8B	INT4 Q8	56 tok/s 32 tok/s	~1.7s ~1.9s
#164	Phi-4	14B	INT4	34 tok/s	~3.0s
#176	DeepSeek-R1 14B	14B	INT4	35 tok/s	~3.0s
#177	Ministral 3 14B	14B	INT4	22 tok/s	~3.6s
#180	Gemma 3 12B	12B	INT4	39 tok/s	~2.6s
#185	Ministral 3 8B	8B	INT4 Q8	40 tok/s 25 tok/s	~1.7s ~2.1s
#187	OLMo 3 7B Think	7B	INT4 Q8	44 tok/s 29 tok/s	~1.5s ~1.7s
#188	Qwen3.5-2B	2B	INT4 Q8 FP16	146 tok/s 102 tok/s 61 tok/s	~480 ms ~483 ms ~489 ms
#189	MiniCPM5-2B	2B	INT4 Q8 FP16	146 tok/s 102 tok/s 61 tok/s	~480 ms ~483 ms ~489 ms
#190	Qwen2.5-14B	14B	INT4	34 tok/s	~3.0s
#193	Gemma 4 E2B	5.1B	INT4 Q8 FP16	80 tok/s 50 tok/s 24 tok/s	~1.1s ~1.1s ~1.4s
#195	Ministral 3 3B	3B	INT4 Q8 FP16	75 tok/s 58 tok/s 37 tok/s	~694 ms ~698 ms ~707 ms
#196	DeepSeek-R1 8B	8B	INT4 Q8	57 tok/s 34 tok/s	~1.7s ~1.8s
#198	Gemma 2 9B	9B	INT4 Q8	50 tok/s 29 tok/s	~2.0s ~2.1s
#201	Qwen3-4B	4B	INT4 Q8 FP16	95 tok/s 61 tok/s 34 tok/s	~900 ms ~906 ms ~919 ms
#202	Qwen3-14B	14B	INT4	34 tok/s	~3.0s
#203	Gemma 3 4B	4B	INT4 Q8 FP16	95 tok/s 61 tok/s 34 tok/s	~900 ms ~906 ms ~919 ms
#204	DeepSeek-R1 7B	7B	INT4 Q8	64 tok/s 38 tok/s	~1.5s ~1.5s
#206	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	229 tok/s 180 tok/s 123 tok/s	~198 ms ~199 ms ~202 ms
#207	Llama 3.1 8B	8B	INT4 Q8	56 tok/s 32 tok/s	~1.7s ~1.9s
#207	Qwen3-8B	8B	INT4 Q8	56 tok/s 32 tok/s	~1.7s ~1.9s
#212	Llama 3 8B	8B	INT4 Q8	56 tok/s 32 tok/s	~1.7s ~1.9s
#213	Qwen3-1.7B	1.7B	INT4 Q8 FP16	160 tok/s 114 tok/s 70 tok/s	~411 ms ~414 ms ~419 ms
#214	Gemma 3 270M	0.27B	INT4 Q8 FP16	273 tok/s 248 tok/s 204 tok/s	~70 ms ~70 ms ~71 ms
#216	Qwen3-0.6B	0.6B	INT4 Q8 FP16	254 tok/s 208 tok/s 148 tok/s	~148 ms ~149 ms ~151 ms
#217	Phi-4-Mini	3.8B	INT4 Q8 FP16	98 tok/s 63 tok/s 35 tok/s	~858 ms ~863 ms ~876 ms
#218	Qwen2.5-7B	7B	INT4 Q8	62 tok/s 38 tok/s	~1.5s ~1.5s
#220	Ministral-8B-2410	8B	INT4 Q8	56 tok/s 32 tok/s	~1.7s ~1.9s
#221	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	171 tok/s 124 tok/s 77 tok/s	~364 ms ~367 ms ~371 ms
#222	Llama 3.2 3B	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
#223	Phi-3-mini	3.8B	INT4 Q8 FP16	98 tok/s 63 tok/s 35 tok/s	~858 ms ~863 ms ~876 ms
#224	Gemma 3n E2B IT	6B (2B active)	INT4 Q8	75 tok/s 64 tok/s	~518 ms ~521 ms
#225	Gemma 2 2B	2B	INT4 Q8 FP16	146 tok/s 102 tok/s 61 tok/s	~480 ms ~483 ms ~489 ms
#226	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8	60 tok/s 36 tok/s	~1.6s ~1.6s
#227	Llama 3.2 1B	1B	INT4 Q8 FP16	207 tok/s 159 tok/s 105 tok/s	~251 ms ~252 ms ~255 ms
#228	Phi-3-medium	14B	INT4	34 tok/s	~3.0s
#230	OLMo 3 7B Instruct	7B	INT4 Q8	44 tok/s 29 tok/s	~1.5s ~1.7s
#231	Qwen2-7B	7B	INT4 Q8	62 tok/s 38 tok/s	~1.5s ~1.5s
#232	Phi-3-small	7B	INT4 Q8	62 tok/s 38 tok/s	~1.5s ~1.5s
#233	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8	60 tok/s 36 tok/s	~1.6s ~1.6s
#234	Gemma 3 1B	1B	INT4 Q8 FP16	207 tok/s 159 tok/s 105 tok/s	~251 ms ~252 ms ~255 ms
#236	Gemma 1 7B	7B	INT4 Q8	44 tok/s 29 tok/s	~1.5s ~1.7s
#237	Gemma 1 2B	2B	INT4 Q8 FP16	156 tok/s 107 tok/s 63 tok/s	~480 ms ~483 ms ~489 ms
#238	ChatGLM3-6B	6B	INT4 Q8	49 tok/s 35 tok/s	~1.3s ~1.3s
#239	ChatGLM2-6B	6B	INT4 Q8	49 tok/s 35 tok/s	~1.3s ~1.3s
#240	ChatGLM-6B	6B	INT4 Q8	49 tok/s 35 tok/s	~1.3s ~1.3s
Unranked Models (Sorted by Speed)
-	DINOv3 ViT-L/16	0.3B	INT4 Q8 FP16	303 tok/s 267 tok/s 212 tok/s	~76 ms ~77 ms ~78 ms
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	269 tok/s 225 tok/s 164 tok/s	~124 ms ~124 ms ~126 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	269 tok/s 225 tok/s 164 tok/s	~124 ms ~124 ms ~126 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	263 tok/s 238 tok/s 193 tok/s	~77 ms ~77 ms ~78 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	263 tok/s 238 tok/s 193 tok/s	~77 ms ~77 ms ~78 ms
-	Liquid D1-Omni 600M	0.6B	INT4 Q8 FP16	254 tok/s 208 tok/s 148 tok/s	~148 ms ~149 ms ~151 ms
-	Falcon-1B	1B	INT4 Q8 FP16	219 tok/s 165 tok/s 107 tok/s	~250 ms ~252 ms ~255 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	207 tok/s 159 tok/s 105 tok/s	~251 ms ~252 ms ~255 ms
-	MiniCPM-V 4.7 1B	1B	INT4 Q8 FP16	207 tok/s 159 tok/s 105 tok/s	~251 ms ~252 ms ~255 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	171 tok/s 124 tok/s 77 tok/s	~364 ms ~367 ms ~371 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	171 tok/s 124 tok/s 77 tok/s	~364 ms ~367 ms ~371 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	160 tok/s 114 tok/s 70 tok/s	~411 ms ~414 ms ~419 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	155 tok/s 110 tok/s 67 tok/s	~433 ms ~435 ms ~441 ms
-	Falcon-3B	3B	INT4 Q8 FP16	122 tok/s 79 tok/s 45 tok/s	~689 ms ~693 ms ~703 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	120 tok/s 100 tok/s 70 tok/s	~321 ms ~323 ms ~327 ms
-	Phi-1	1.3B	INT4 Q8 FP16	120 tok/s 100 tok/s 70 tok/s	~321 ms ~323 ms ~327 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	120 tok/s 100 tok/s 70 tok/s	~321 ms ~323 ms ~327 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	117 tok/s 77 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	Mellum 2.1 12B A2.5B Thinking	12B (2.5B active)	INT4	117 tok/s	~673 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	LFM2-VL 3B	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	Liquid D1 3B	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	Liquid D1-VL 3B	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	Liquid LFM2-VL 3B D1	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	115 tok/s 76 tok/s 44 tok/s	~689 ms ~693 ms ~703 ms
-	FrogNano 4B	4B	INT4 Q8 FP16	95 tok/s 61 tok/s 34 tok/s	~900 ms ~906 ms ~919 ms
-	Spark X2.5 4B	4B	INT4 Q8 FP16	95 tok/s 61 tok/s 34 tok/s	~900 ms ~906 ms ~919 ms
-	EmbeddingGemma 2	2B	INT4 Q8 FP16	91 tok/s 73 tok/s 50 tok/s	~484 ms ~487 ms ~493 ms
-	Phi-2	2.7B	INT4 Q8 FP16	79 tok/s 62 tok/s 40 tok/s	~631 ms ~634 ms ~643 ms
-	MaLLaM-3B	3B	INT4 Q8 FP16	75 tok/s 58 tok/s 37 tok/s	~694 ms ~698 ms ~707 ms
-	SmolLM3 3B	3B	INT4 Q8 FP16	75 tok/s 58 tok/s 37 tok/s	~694 ms ~698 ms ~707 ms
-	Falcon-7B	7B	INT4 Q8	65 tok/s 39 tok/s	~1.5s ~1.5s
-	Falcon3-7B	7B	INT4 Q8	62 tok/s 38 tok/s	~1.5s ~1.5s
-	Hy-MT2 7B	7B	INT4 Q8	62 tok/s 38 tok/s	~1.5s ~1.5s
-	Mistral-7B-v0.1	7.3B	INT4 Q8	60 tok/s 36 tok/s	~1.6s ~1.6s
-	Youtu Parsing Omni	8B	INT4 Q8	57 tok/s 34 tok/s	~1.7s ~1.8s
-	LensVLM 9B	9B	INT4 Q8	50 tok/s 29 tok/s	~2.0s ~2.1s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8	50 tok/s 29 tok/s	~2.0s ~2.1s
-	ChatGLM3-6B-32K	6B	INT4 Q8	49 tok/s 35 tok/s	~1.3s ~1.3s
-	Yi-6B	6B	INT4 Q8	49 tok/s 35 tok/s	~1.3s ~1.3s
-	Falcon3-10B	10B	INT4 Q8	46 tok/s 25 tok/s	~2.2s ~2.6s
-	Falcon2-11B	11B	INT4 Q8	44 tok/s 23 tok/s	~2.4s ~2.8s
-	Hunyuan Lite	7B	INT4 Q8	44 tok/s 29 tok/s	~1.5s ~1.7s
-	OLMo 3 7B Base	7B	INT4 Q8	44 tok/s 29 tok/s	~1.5s ~1.7s
-	SEA-LION-7B	7.1B	INT4 Q8	44 tok/s 29 tok/s	~1.6s ~1.7s
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8	44 tok/s 29 tok/s	~1.6s ~1.7s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8	40 tok/s 25 tok/s	~1.7s ~2.1s
-	Typhoon-2-8B	8B	INT4 Q8	40 tok/s 25 tok/s	~1.7s ~2.1s
-	GLM-4-9B	9B	INT4	36 tok/s	~2.0s
-	GLM-4-9B-Chat	9B	INT4	36 tok/s	~2.0s
-	GLM-4-9B-Chat-1M	9B	INT4	36 tok/s	~2.0s
-	Sahabat-AI-Gemma2-9B	9.2B	INT4	36 tok/s	~2.0s
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4	36 tok/s	~2.0s
-	Yi-9B	9B	INT4	36 tok/s	~2.0s
-	GreenMind-14B-R1	14B	INT4	22 tok/s	~3.6s

NVIDIA RTX 3060 (12GB)

VRAM Scaling by Context Length

Best Models to Run on NVIDIA RTX 3060 (12GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions