What are the best models to run on AMD RX 7700 XT (12GB)?

Based on model rankings and 12 GB usable VRAM: Top Overall: Phi-4 Reasoning Plus (INT4) ~27 tok/s. Top Coding: K2 Horizon 7B (INT4) ~72 tok/s. Optimal Daily Driver: Qwen3.5-9B (INT4) ~59 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 AMD RX 7700 XT (12GB)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#102	Phi-4 Reasoning Plus	14B	INT4	27 tok/s	~2.6s
#121	Qwen3.5-9B	9B	INT4 Q8	59 tok/s 34 tok/s	~1.4s ~1.6s
#127	K2 Horizon 7B	7B	INT4 Q8	72 tok/s 44 tok/s	~1.1s ~1.1s
#130	Gemma 4 12B	11.95B	INT4	32 tok/s	~2.1s
#136	Qwen3.5-4B	4B	INT4 Q8 FP16	108 tok/s 71 tok/s 40 tok/s	~661 ms ~666 ms ~677 ms
#140	GLM-4V	9B	INT4	42 tok/s	~1.4s
#145	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	19 tok/s 11 tok/s 6 tok/s	~4.5s ~4.6s ~5.0s
#152	Granite 4.2 8B	8B	INT4 Q8	65 tok/s 38 tok/s	~1.3s ~1.4s
#157	Gemma 4 E4B	8B	INT4 Q8	65 tok/s 38 tok/s	~1.3s ~1.4s
#164	Phi-4	14B	INT4	40 tok/s	~2.2s
#176	DeepSeek-R1 14B	14B	INT4	42 tok/s	~2.2s
#177	Ministral 3 14B	14B	INT4	27 tok/s	~2.6s
#180	Gemma 3 12B	12B	INT4	46 tok/s	~1.9s
#185	Ministral 3 8B	8B	INT4 Q8	47 tok/s 29 tok/s	~1.3s ~1.5s
#187	OLMo 3 7B Think	7B	INT4 Q8	52 tok/s 34 tok/s	~1.1s ~1.2s
#188	Qwen3.5-2B	2B	INT4 Q8 FP16	162 tok/s 116 tok/s 71 tok/s	~354 ms ~356 ms ~362 ms
#189	MiniCPM5-2B	2B	INT4 Q8 FP16	162 tok/s 116 tok/s 71 tok/s	~354 ms ~356 ms ~362 ms
#190	Qwen2.5-14B	14B	INT4	40 tok/s	~2.2s
#193	Gemma 4 E2B	5.1B	INT4 Q8 FP16	92 tok/s 58 tok/s 29 tok/s	~831 ms ~838 ms ~999 ms
#195	Ministral 3 3B	3B	INT4 Q8 FP16	87 tok/s 67 tok/s 43 tok/s	~511 ms ~514 ms ~522 ms
#196	DeepSeek-R1 8B	8B	INT4 Q8	67 tok/s 40 tok/s	~1.3s ~1.3s
#198	Gemma 2 9B	9B	INT4 Q8	59 tok/s 34 tok/s	~1.4s ~1.6s
#201	Qwen3-4B	4B	INT4 Q8 FP16	108 tok/s 71 tok/s 40 tok/s	~661 ms ~666 ms ~677 ms
#202	Qwen3-14B	14B	INT4	40 tok/s	~2.2s
#203	Gemma 3 4B	4B	INT4 Q8 FP16	108 tok/s 71 tok/s 40 tok/s	~661 ms ~666 ms ~677 ms
#204	DeepSeek-R1 7B	7B	INT4 Q8	74 tok/s 45 tok/s	~1.1s ~1.1s
#206	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	244 tok/s 197 tok/s 138 tok/s	~147 ms ~148 ms ~151 ms
#207	Llama 3.1 8B	8B	INT4 Q8	65 tok/s 38 tok/s	~1.3s ~1.4s
#207	Qwen3-8B	8B	INT4 Q8	65 tok/s 38 tok/s	~1.3s ~1.4s
#212	Llama 3 8B	8B	INT4 Q8	65 tok/s 38 tok/s	~1.3s ~1.4s
#213	Qwen3-1.7B	1.7B	INT4 Q8 FP16	177 tok/s 129 tok/s 81 tok/s	~303 ms ~305 ms ~310 ms
#214	Gemma 3 270M	0.27B	INT4 Q8 FP16	285 tok/s 263 tok/s 220 tok/s	~53 ms ~53 ms ~54 ms
#216	Qwen3-0.6B	0.6B	INT4 Q8 FP16	268 tok/s 224 tok/s 164 tok/s	~111 ms ~111 ms ~113 ms
#217	Phi-4-Mini	3.8B	INT4 Q8 FP16	112 tok/s 74 tok/s 42 tok/s	~631 ms ~635 ms ~646 ms
#218	Qwen2.5-7B	7B	INT4 Q8	72 tok/s 44 tok/s	~1.1s ~1.1s
#220	Ministral-8B-2410	8B	INT4 Q8	65 tok/s 38 tok/s	~1.3s ~1.4s
#221	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	188 tok/s 140 tok/s 89 tok/s	~269 ms ~271 ms ~275 ms
#222	Llama 3.2 3B	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
#223	Phi-3-mini	3.8B	INT4 Q8 FP16	112 tok/s 74 tok/s 42 tok/s	~631 ms ~635 ms ~646 ms
#224	Gemma 3n E2B IT	6B (2B active)	INT4 Q8	87 tok/s 74 tok/s	~382 ms ~384 ms
#225	Gemma 2 2B	2B	INT4 Q8 FP16	162 tok/s 116 tok/s 71 tok/s	~354 ms ~356 ms ~362 ms
#226	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8	70 tok/s 43 tok/s	~1.2s ~1.2s
#227	Llama 3.2 1B	1B	INT4 Q8 FP16	223 tok/s 176 tok/s 119 tok/s	~186 ms ~187 ms ~190 ms
#228	Phi-3-medium	14B	INT4	40 tok/s	~2.2s
#230	OLMo 3 7B Instruct	7B	INT4 Q8	52 tok/s 34 tok/s	~1.1s ~1.2s
#231	Qwen2-7B	7B	INT4 Q8	72 tok/s 44 tok/s	~1.1s ~1.1s
#232	Phi-3-small	7B	INT4 Q8	72 tok/s 44 tok/s	~1.1s ~1.1s
#233	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8	70 tok/s 43 tok/s	~1.2s ~1.2s
#234	Gemma 3 1B	1B	INT4 Q8 FP16	223 tok/s 176 tok/s 119 tok/s	~186 ms ~187 ms ~190 ms
#236	Gemma 1 7B	7B	INT4 Q8	52 tok/s 34 tok/s	~1.1s ~1.2s
#237	Gemma 1 2B	2B	INT4 Q8 FP16	173 tok/s 121 tok/s 73 tok/s	~353 ms ~356 ms ~361 ms
#238	ChatGLM3-6B	6B	INT4 Q8	57 tok/s 41 tok/s	~976 ms ~983 ms
#239	ChatGLM2-6B	6B	INT4 Q8	57 tok/s 41 tok/s	~976 ms ~983 ms
#240	ChatGLM-6B	6B	INT4 Q8	57 tok/s 41 tok/s	~976 ms ~983 ms
Unranked Models (Sorted by Speed)
-	DINOv3 ViT-L/16	0.3B	INT4 Q8 FP16	312 tok/s 280 tok/s 228 tok/s	~58 ms ~58 ms ~59 ms
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	282 tok/s 240 tok/s 181 tok/s	~93 ms ~93 ms ~95 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	282 tok/s 240 tok/s 181 tok/s	~93 ms ~93 ms ~95 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	277 tok/s 253 tok/s 210 tok/s	~58 ms ~59 ms ~60 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	277 tok/s 253 tok/s 210 tok/s	~58 ms ~59 ms ~60 ms
-	Liquid D1-Omni 600M	0.6B	INT4 Q8 FP16	268 tok/s 224 tok/s 164 tok/s	~111 ms ~111 ms ~113 ms
-	Falcon-1B	1B	INT4 Q8 FP16	235 tok/s 182 tok/s 122 tok/s	~186 ms ~187 ms ~189 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	223 tok/s 176 tok/s 119 tok/s	~186 ms ~187 ms ~190 ms
-	MiniCPM-V 4.7 1B	1B	INT4 Q8 FP16	223 tok/s 176 tok/s 119 tok/s	~186 ms ~187 ms ~190 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	188 tok/s 140 tok/s 89 tok/s	~269 ms ~271 ms ~275 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	188 tok/s 140 tok/s 89 tok/s	~269 ms ~271 ms ~275 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	177 tok/s 129 tok/s 81 tok/s	~303 ms ~305 ms ~310 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	172 tok/s 125 tok/s 77 tok/s	~319 ms ~321 ms ~326 ms
-	Falcon-3B	3B	INT4 Q8 FP16	138 tok/s 91 tok/s 52 tok/s	~507 ms ~510 ms ~519 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	135 tok/s 114 tok/s 81 tok/s	~238 ms ~239 ms ~242 ms
-	Phi-1	1.3B	INT4 Q8 FP16	135 tok/s 114 tok/s 81 tok/s	~238 ms ~239 ms ~242 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	135 tok/s 114 tok/s 81 tok/s	~238 ms ~239 ms ~242 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	132 tok/s 89 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	Mellum 2.1 12B A2.5B Thinking	12B (2.5B active)	INT4	132 tok/s	~495 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	LFM2-VL 3B	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	Liquid D1 3B	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	Liquid D1-VL 3B	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	Liquid LFM2-VL 3B D1	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	130 tok/s 88 tok/s 51 tok/s	~507 ms ~511 ms ~519 ms
-	FrogNano 4B	4B	INT4 Q8 FP16	108 tok/s 71 tok/s 40 tok/s	~661 ms ~666 ms ~677 ms
-	Spark X2.5 4B	4B	INT4 Q8 FP16	108 tok/s 71 tok/s 40 tok/s	~661 ms ~666 ms ~677 ms
-	EmbeddingGemma 2	2B	INT4 Q8 FP16	104 tok/s 85 tok/s 58 tok/s	~357 ms ~359 ms ~365 ms
-	Phi-2	2.7B	INT4 Q8 FP16	91 tok/s 72 tok/s 47 tok/s	~465 ms ~468 ms ~475 ms
-	MaLLaM-3B	3B	INT4 Q8 FP16	87 tok/s 67 tok/s 43 tok/s	~511 ms ~514 ms ~522 ms
-	SmolLM3 3B	3B	INT4 Q8 FP16	87 tok/s 67 tok/s 43 tok/s	~511 ms ~514 ms ~522 ms
-	Falcon-7B	7B	INT4 Q8	75 tok/s 46 tok/s	~1.1s ~1.1s
-	Falcon3-7B	7B	INT4 Q8	72 tok/s 44 tok/s	~1.1s ~1.1s
-	Hy-MT2 7B	7B	INT4 Q8	72 tok/s 44 tok/s	~1.1s ~1.1s
-	Mistral-7B-v0.1	7.3B	INT4 Q8	70 tok/s 43 tok/s	~1.2s ~1.2s
-	Youtu Parsing Omni	8B	INT4 Q8	67 tok/s 40 tok/s	~1.3s ~1.3s
-	LensVLM 9B	9B	INT4 Q8	59 tok/s 34 tok/s	~1.4s ~1.6s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8	59 tok/s 34 tok/s	~1.4s ~1.6s
-	ChatGLM3-6B-32K	6B	INT4 Q8	57 tok/s 41 tok/s	~976 ms ~983 ms
-	Yi-6B	6B	INT4 Q8	57 tok/s 41 tok/s	~976 ms ~983 ms
-	Falcon3-10B	10B	INT4 Q8	54 tok/s 29 tok/s	~1.6s ~1.9s
-	Falcon2-11B	11B	INT4 Q8	52 tok/s 27 tok/s	~1.7s ~2.1s
-	Hunyuan Lite	7B	INT4 Q8	52 tok/s 34 tok/s	~1.1s ~1.2s
-	OLMo 3 7B Base	7B	INT4 Q8	52 tok/s 34 tok/s	~1.1s ~1.2s
-	SEA-LION-7B	7.1B	INT4 Q8	51 tok/s 34 tok/s	~1.1s ~1.3s
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8	51 tok/s 34 tok/s	~1.1s ~1.3s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8	47 tok/s 29 tok/s	~1.3s ~1.5s
-	Typhoon-2-8B	8B	INT4 Q8	47 tok/s 29 tok/s	~1.3s ~1.5s
-	GLM-4-9B	9B	INT4	42 tok/s	~1.4s
-	GLM-4-9B-Chat	9B	INT4	42 tok/s	~1.4s
-	GLM-4-9B-Chat-1M	9B	INT4	42 tok/s	~1.4s
-	Sahabat-AI-Gemma2-9B	9.2B	INT4	42 tok/s	~1.5s
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4	42 tok/s	~1.5s
-	Yi-9B	9B	INT4	42 tok/s	~1.4s
-	GreenMind-14B-R1	14B	INT4	27 tok/s	~2.6s

AMD RX 7700 XT (12GB)

VRAM Scaling by Context Length

Best Models to Run on AMD RX 7700 XT (12GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions