What are the best models to run on NVIDIA RTX 4070 (12GB)?

Based on model rankings and 12 GB usable VRAM: Top Overall: Phi-4 Reasoning Plus (INT4) ~31 tok/s. Top Coding: K2 Horizon 7B (INT4) ~82 tok/s. Optimal Daily Driver: Gemma 4 12B (INT4) ~37 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 4070 (12GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#100	Phi-4 Reasoning Plus	14B	INT4	31 tok/s	~1.6s
#109	Gemma 4 12B	11.95B	INT4	37 tok/s	~1.3s
#113	Qwen3.5-9B	9B	INT4 Q8	67 tok/s 39 tok/s	~871 ms ~955 ms
#118	K2 Horizon 7B	7B	INT4 Q8	82 tok/s 51 tok/s	~683 ms ~690 ms
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	22 tok/s 13 tok/s 6 tok/s	~2.7s ~2.8s ~3.1s
#132	Qwen3.5-4B	4B	INT4 Q8 FP16	121 tok/s 80 tok/s 46 tok/s	~403 ms ~407 ms ~416 ms
#146	Granite 4.2 8B	8B	INT4 Q8	74 tok/s 43 tok/s	~776 ms ~851 ms
#150	Gemma 4 E4B	8B	INT4 Q8	74 tok/s 43 tok/s	~776 ms ~851 ms
#158	GLM-4V	9B	INT4	49 tok/s	~876 ms
#160	Phi-4	14B	INT4	46 tok/s	~1.3s
#169	OLMo 3 7B Think	7B	INT4 Q8	59 tok/s 39 tok/s	~687 ms ~753 ms
#172	DeepSeek-R1 14B	14B	INT4	48 tok/s	~1.3s
#173	Ministral 3 14B	14B	INT4	31 tok/s	~1.6s
#177	Gemma 3 12B	12B	INT4	53 tok/s	~1.2s
#182	Ministral 3 8B	8B	INT4 Q8	53 tok/s 34 tok/s	~781 ms ~925 ms
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	177 tok/s 129 tok/s 80 tok/s	~217 ms ~219 ms ~224 ms
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	177 tok/s 129 tok/s 80 tok/s	~217 ms ~219 ms ~224 ms
#185	Qwen2.5-14B	14B	INT4	46 tok/s	~1.3s
#188	Gemma 4 E2B	5.1B	INT4 Q8 FP16	103 tok/s 66 tok/s 33 tok/s	~505 ms ~511 ms ~612 ms
#189	Qwen3-4B	4B	INT4 Q8 FP16	121 tok/s 80 tok/s 46 tok/s	~403 ms ~407 ms ~416 ms
#190	DeepSeek-R1 8B	8B	INT4 Q8	76 tok/s 46 tok/s	~775 ms ~784 ms
#193	Ministral 3 3B	3B	INT4 Q8 FP16	98 tok/s 76 tok/s 50 tok/s	~313 ms ~315 ms ~323 ms
#195	Qwen3-14B	14B	INT4	46 tok/s	~1.3s
#197	Gemma 2 9B	9B	INT4 Q8	67 tok/s 39 tok/s	~871 ms ~955 ms
#198	Qwen2.5-7B	7B	INT4 Q8	82 tok/s 51 tok/s	~683 ms ~690 ms
#199	Qwen3-8B	8B	INT4 Q8	74 tok/s 43 tok/s	~776 ms ~851 ms
#200	Gemma 3 4B	4B	INT4 Q8 FP16	121 tok/s 80 tok/s 46 tok/s	~403 ms ~407 ms ~416 ms
#201	DeepSeek-R1 7B	7B	INT4 Q8	84 tok/s 52 tok/s	~682 ms ~690 ms
#202	Llama 3.1 8B	8B	INT4 Q8	74 tok/s 43 tok/s	~776 ms ~851 ms
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	191 tok/s 143 tok/s 91 tok/s	~186 ms ~188 ms ~192 ms
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	295 tok/s 274 tok/s 234 tok/s	~35 ms ~35 ms ~36 ms
#208	Llama 3 8B	8B	INT4 Q8	74 tok/s 43 tok/s	~776 ms ~851 ms
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	279 tok/s 238 tok/s 178 tok/s	~70 ms ~70 ms ~72 ms
#210	Phi-4-Mini	3.8B	INT4 Q8 FP16	125 tok/s 83 tok/s 48 tok/s	~384 ms ~388 ms ~397 ms
#212	Ministral-8B-2410	8B	INT4 Q8	74 tok/s 43 tok/s	~776 ms ~851 ms
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	203 tok/s 154 tok/s 100 tok/s	~166 ms ~167 ms ~171 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	143 tok/s 99 tok/s 58 tok/s	~309 ms ~312 ms ~319 ms
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	257 tok/s 211 tok/s 151 tok/s	~92 ms ~93 ms ~95 ms
#216	Phi-3-mini	3.8B	INT4 Q8 FP16	125 tok/s 83 tok/s 48 tok/s	~384 ms ~388 ms ~397 ms
#217	Gemma 3n E2B IT	6B (2B active)	INT4 Q8	97 tok/s 83 tok/s	~235 ms ~237 ms
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8	79 tok/s 49 tok/s	~710 ms ~718 ms
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	237 tok/s 190 tok/s 132 tok/s	~115 ms ~116 ms ~119 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	177 tok/s 129 tok/s 80 tok/s	~217 ms ~219 ms ~224 ms
#221	Phi-3-medium	14B	INT4	46 tok/s	~1.3s
#222	OLMo 3 7B Instruct	7B	INT4 Q8	59 tok/s 39 tok/s	~687 ms ~753 ms
#224	Qwen2-7B	7B	INT4 Q8	82 tok/s 51 tok/s	~683 ms ~690 ms
#225	Phi-3-small	7B	INT4 Q8	82 tok/s 51 tok/s	~683 ms ~690 ms
#226	Gemma 3 1B	1B	INT4 Q8 FP16	237 tok/s 190 tok/s 132 tok/s	~115 ms ~116 ms ~119 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8	79 tok/s 49 tok/s	~710 ms ~718 ms
#229	Gemma 1 7B	7B	INT4 Q8	59 tok/s 39 tok/s	~687 ms ~753 ms
#230	Gemma 1 2B	2B	INT4 Q8 FP16	187 tok/s 134 tok/s 82 tok/s	~216 ms ~219 ms ~223 ms
#231	ChatGLM3-6B	6B	INT4 Q8	65 tok/s 47 tok/s	~594 ms ~600 ms
#232	ChatGLM2-6B	6B	INT4 Q8	65 tok/s 47 tok/s	~594 ms ~600 ms
#233	ChatGLM-6B	6B	INT4 Q8	65 tok/s 47 tok/s	~594 ms ~600 ms
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	292 tok/s 253 tok/s 195 tok/s	~59 ms ~60 ms ~61 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	292 tok/s 253 tok/s 195 tok/s	~59 ms ~60 ms ~61 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	287 tok/s 265 tok/s 224 tok/s	~38 ms ~39 ms ~39 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	287 tok/s 265 tok/s 224 tok/s	~38 ms ~39 ms ~39 ms
-	Falcon-1B	1B	INT4 Q8 FP16	248 tok/s 196 tok/s 135 tok/s	~115 ms ~116 ms ~118 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	237 tok/s 190 tok/s 132 tok/s	~115 ms ~116 ms ~119 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	203 tok/s 154 tok/s 100 tok/s	~166 ms ~167 ms ~171 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	203 tok/s 154 tok/s 100 tok/s	~166 ms ~167 ms ~171 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	191 tok/s 143 tok/s 91 tok/s	~186 ms ~188 ms ~192 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	186 tok/s 138 tok/s 87 tok/s	~196 ms ~198 ms ~202 ms
-	Falcon-3B	3B	INT4 Q8 FP16	151 tok/s 102 tok/s 59 tok/s	~309 ms ~312 ms ~319 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	149 tok/s 126 tok/s 91 tok/s	~147 ms ~148 ms ~151 ms
-	Phi-1	1.3B	INT4 Q8 FP16	149 tok/s 126 tok/s 91 tok/s	~147 ms ~148 ms ~151 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	149 tok/s 126 tok/s 91 tok/s	~147 ms ~148 ms ~151 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	145 tok/s 99 tok/s 59 tok/s	~309 ms ~312 ms ~319 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	143 tok/s 99 tok/s 58 tok/s	~309 ms ~312 ms ~319 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	143 tok/s 99 tok/s 58 tok/s	~309 ms ~312 ms ~319 ms
-	Spark X2.5 4B	4B	INT4 Q8 FP16	121 tok/s 80 tok/s 46 tok/s	~403 ms ~407 ms ~416 ms
-	Phi-2	2.7B	INT4 Q8 FP16	102 tok/s 81 tok/s 54 tok/s	~285 ms ~287 ms ~294 ms
-	MaLLaM-3B	3B	INT4 Q8 FP16	98 tok/s 76 tok/s 50 tok/s	~313 ms ~315 ms ~323 ms
-	SmolLM3 3B	3B	INT4 Q8 FP16	98 tok/s 76 tok/s 50 tok/s	~313 ms ~315 ms ~323 ms
-	Falcon-7B	7B	INT4 Q8	85 tok/s 52 tok/s	~682 ms ~690 ms
-	Falcon3-7B	7B	INT4 Q8	82 tok/s 51 tok/s	~683 ms ~690 ms
-	Hy-MT2 7B	7B	INT4 Q8	82 tok/s 51 tok/s	~683 ms ~690 ms
-	Mistral-7B-v0.1	7.3B	INT4 Q8	79 tok/s 49 tok/s	~710 ms ~718 ms
-	LensVLM 9B	9B	INT4 Q8	67 tok/s 39 tok/s	~871 ms ~955 ms
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8	67 tok/s 39 tok/s	~871 ms ~955 ms
-	ChatGLM3-6B-32K	6B	INT4 Q8	65 tok/s 47 tok/s	~594 ms ~600 ms
-	Yi-6B	6B	INT4 Q8	65 tok/s 47 tok/s	~594 ms ~600 ms
-	Falcon3-10B	10B	INT4 Q8	62 tok/s 34 tok/s	~964 ms ~1.1s
-	Falcon2-11B	11B	INT4 Q8	59 tok/s 31 tok/s	~1.1s ~1.3s
-	Hunyuan Lite	7B	INT4 Q8	59 tok/s 39 tok/s	~687 ms ~753 ms
-	OLMo 3 7B Base	7B	INT4 Q8	59 tok/s 39 tok/s	~687 ms ~753 ms
-	SEA-LION-7B	7.1B	INT4 Q8	58 tok/s 39 tok/s	~697 ms ~763 ms
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8	58 tok/s 39 tok/s	~697 ms ~763 ms
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8	53 tok/s 34 tok/s	~781 ms ~925 ms
-	Typhoon-2-8B	8B	INT4 Q8	53 tok/s 34 tok/s	~781 ms ~925 ms
-	GLM-4-9B	9B	INT4	49 tok/s	~876 ms
-	GLM-4-9B-Chat	9B	INT4	49 tok/s	~876 ms
-	GLM-4-9B-Chat-1M	9B	INT4	49 tok/s	~876 ms
-	Yi-9B	9B	INT4	49 tok/s	~876 ms
-	Sahabat-AI-Gemma2-9B	9.2B	INT4	48 tok/s	~895 ms
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4	48 tok/s	~895 ms
-	GreenMind-14B-R1	14B	INT4	31 tok/s	~1.6s

NVIDIA RTX 4070 (12GB)

VRAM Scaling by Context Length

Best Models to Run on NVIDIA RTX 4070 (12GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions