What are the best models to run on AMD RX 7600 (8GB)?

Based on model rankings and 8 GB usable VRAM: Top Overall: Qwen3.5-9B (INT4) ~40 tok/s. Top Coding: K2 Horizon 7B (INT4) ~49 tok/s. Optimal Daily Driver: Granite 4.2 8B (INT4) ~44 tok/s. Maximum Model Size: HunyuanVideo 1.5 720P T2V (INT4) ~26 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 7600 (8GB)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#125	Qwen3.5-9B	9B	INT4	40 tok/s	~2.7s
#128	K2 Horizon 7B	7B	INT4	49 tok/s	~2.1s
#140	Qwen3.5-4B	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
#150	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8	13 tok/s 7 tok/s	~8.4s ~9.2s
#158	Granite 4.2 8B	8B	INT4	44 tok/s	~2.4s
#159	Gemma 4 E4B	8B	INT4	44 tok/s	~2.4s
#178	OLMo 3 7B Think	7B	INT4	33 tok/s	~2.3s
#186	Gemma 3 12B	12B	INT4	28 tok/s	~4.1s
#189	Ministral 3 8B	8B	INT4	29 tok/s	~2.8s
#191	MiniCPM5-2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
#192	Qwen3.5-2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
#200	Gemma 4 E2B	5.1B	INT4 Q8	63 tok/s 38 tok/s	~1.5s ~1.7s
#202	DeepSeek-R1 8B	8B	INT4	46 tok/s	~2.4s
#204	Qwen3-4B	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
#205	Ministral 3 3B	3B	INT4 Q8	60 tok/s 46 tok/s	~941 ms ~946 ms
#207	Gemma 2 9B	9B	INT4	40 tok/s	~2.7s
#208	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	179 tok/s 142 tok/s 97 tok/s	~269 ms ~270 ms ~273 ms
#210	Gemma 3 4B	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
#211	Qwen3-8B	8B	INT4	44 tok/s	~2.4s
#213	DeepSeek-R1 7B	7B	INT4 Q8	51 tok/s 28 tok/s	~2.1s ~2.5s
#214	Qwen3-1.7B	1.7B	INT4 Q8 FP16	126 tok/s 91 tok/s 56 tok/s	~558 ms ~561 ms ~568 ms
#215	Llama 3.1 8B	8B	INT4	44 tok/s	~2.4s
#216	Gemma 3 270M	0.27B	INT4 Q8 FP16	212 tok/s 194 tok/s 160 tok/s	~94 ms ~94 ms ~95 ms
#218	Qwen3-0.6B	0.6B	INT4 Q8 FP16	198 tok/s 163 tok/s 116 tok/s	~200 ms ~201 ms ~204 ms
#219	Llama 3 8B	8B	INT4	44 tok/s	~2.4s
#221	Phi-4-Mini	3.8B	INT4 Q8	78 tok/s 50 tok/s	~1.2s ~1.2s
#223	Qwen2.5-7B	7B	INT4	49 tok/s	~2.1s
#224	Ministral-8B-2410	8B	INT4	44 tok/s	~2.4s
#225	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	135 tok/s 98 tok/s 61 tok/s	~494 ms ~497 ms ~503 ms
#226	Llama 3.2 3B	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
#228	Phi-3-mini	3.8B	INT4 Q8	78 tok/s 50 tok/s	~1.2s ~1.2s
#229	Gemma 3n E2B IT	6B (2B active)	INT4	57 tok/s	~763 ms
#230	Mistral-7B-Instruct-v0.1	7.3B	INT4	48 tok/s	~2.2s
#231	Llama 3.2 1B	1B	INT4 Q8 FP16	162 tok/s 125 tok/s 83 tok/s	~340 ms ~341 ms ~346 ms
#232	Gemma 2 2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
#234	OLMo 3 7B Instruct	7B	INT4	33 tok/s	~2.3s
#236	Qwen2-7B	7B	INT4	49 tok/s	~2.1s
#237	Phi-3-small	7B	INT4	49 tok/s	~2.1s
#238	Gemma 3 1B	1B	INT4 Q8 FP16	162 tok/s 125 tok/s 83 tok/s	~340 ms ~341 ms ~346 ms
#239	Mistral-7B-Instruct-v0.2	7.3B	INT4	48 tok/s	~2.2s
#241	Gemma 1 7B	7B	INT4	33 tok/s	~2.3s
#242	Gemma 1 2B	2B	INT4 Q8 FP16	123 tok/s 84 tok/s 50 tok/s	~651 ms ~654 ms ~663 ms
#243	ChatGLM3-6B	6B	INT4	37 tok/s	~2.0s
#244	ChatGLM2-6B	6B	INT4	37 tok/s	~2.0s
#245	ChatGLM-6B	6B	INT4	37 tok/s	~2.0s
Unranked Models (Sorted by Speed)
-	DINOv3 ViT-L/16	0.3B	INT4 Q8 FP16	235 tok/s 208 tok/s 166 tok/s	~103 ms ~103 ms ~105 ms
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	210 tok/s 176 tok/s 129 tok/s	~167 ms ~168 ms ~170 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	210 tok/s 176 tok/s 129 tok/s	~167 ms ~168 ms ~170 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	205 tok/s 186 tok/s 151 tok/s	~104 ms ~104 ms ~105 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	205 tok/s 186 tok/s 151 tok/s	~104 ms ~104 ms ~105 ms
-	Liquid D1-Omni 600M	0.6B	INT4 Q8 FP16	198 tok/s 163 tok/s 116 tok/s	~200 ms ~201 ms ~204 ms
-	PixArt-α XL 2 1024-MS	0.61B	INT4 Q8 FP16	197 tok/s 162 tok/s 115 tok/s	~203 ms ~204 ms ~207 ms
-	PixArt-Σ XL 2 1024-MS	0.61B	INT4 Q8 FP16	197 tok/s 162 tok/s 115 tok/s	~203 ms ~204 ms ~207 ms
-	Nemotron 3.5 Parse	0.63B	INT4 Q8 FP16	195 tok/s 159 tok/s 113 tok/s	~211 ms ~212 ms ~215 ms
-	SD 2.1 Turbo	0.86B	INT4 Q8 FP16	174 tok/s 136 tok/s 92 tok/s	~289 ms ~291 ms ~294 ms
-	Stable Diffusion 1.4	0.86B	INT4 Q8 FP16	174 tok/s 136 tok/s 92 tok/s	~289 ms ~291 ms ~294 ms
-	Stable Diffusion 2.1	0.86B	INT4 Q8 FP16	174 tok/s 136 tok/s 92 tok/s	~289 ms ~291 ms ~294 ms
-	Falcon-1B	1B	INT4 Q8 FP16	172 tok/s 130 tok/s 85 tok/s	~339 ms ~341 ms ~345 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	162 tok/s 125 tok/s 83 tok/s	~340 ms ~341 ms ~346 ms
-	MiniCPM-V 4.7 1B	1B	INT4 Q8 FP16	162 tok/s 125 tok/s 83 tok/s	~340 ms ~341 ms ~346 ms
-	Kandinsky 2.2 Decoder	1.22B	INT4 Q8 FP16	149 tok/s 112 tok/s 72 tok/s	~408 ms ~410 ms ~415 ms
-	Wan 2.1 T2V 1.3B	1.3B	INT4 Q8 FP16	145 tok/s 108 tok/s 68 tok/s	~431 ms ~434 ms ~439 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	135 tok/s 98 tok/s 61 tok/s	~494 ms ~497 ms ~503 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	135 tok/s 98 tok/s 61 tok/s	~494 ms ~497 ms ~503 ms
-	SVD Image-to-Video XT	1.5B	INT4 Q8 FP16	135 tok/s 98 tok/s 61 tok/s	~494 ms ~497 ms ~503 ms
-	SVD Image-to-Video XT 1.1	1.5B	INT4 Q8 FP16	135 tok/s 98 tok/s 61 tok/s	~494 ms ~497 ms ~503 ms
-	LFM3 1.7B	1.7B	INT4 Q8 FP16	126 tok/s 91 tok/s 56 tok/s	~558 ms ~561 ms ~568 ms
-	LFM3 1.7B Base	1.7B	INT4 Q8 FP16	126 tok/s 91 tok/s 56 tok/s	~558 ms ~561 ms ~568 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	126 tok/s 91 tok/s 56 tok/s	~558 ms ~561 ms ~568 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	122 tok/s 87 tok/s 53 tok/s	~587 ms ~590 ms ~598 ms
-	CogVideoX 2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	LFM3 Audio 2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	LFM3 Omni 2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	LFM3 VL 2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	Liquid D1 2B	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	LTX-Video 0.9.1	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	Moondream 3.1	9B (2B active)	INT4	115 tok/s	~733 ms
-	Stable Diffusion 3 Medium	2B	INT4 Q8 FP16	115 tok/s 81 tok/s 49 tok/s	~651 ms ~655 ms ~663 ms
-	Kandinsky 3.0	2.1B	INT4 Q8 FP16	112 tok/s 78 tok/s 47 tok/s	~680 ms ~684 ms ~692 ms
-	LTX-2	2.2B	INT4 Q8 FP16	109 tok/s 76 tok/s 45 tok/s	~708 ms ~712 ms ~721 ms
-	SDXL 1.0 Refiner	2.31B	INT4 Q8 FP16	106 tok/s 73 tok/s 43 tok/s	~739 ms ~744 ms ~753 ms
-	Stable Diffusion 3.5 Medium	2.5B	INT4 Q8 FP16	102 tok/s 69 tok/s 41 tok/s	~793 ms ~798 ms ~808 ms
-	SDXL 1.0 Base	2.56B	INT4 Q8 FP16	100 tok/s 68 tok/s 38 tok/s	~810 ms ~815 ms ~896 ms
-	SDXL Turbo	2.56B	INT4 Q8 FP16	100 tok/s 68 tok/s 38 tok/s	~810 ms ~815 ms ~896 ms
-	Falcon-3B	3B	INT4 Q8 FP16	97 tok/s 63 tok/s 34 tok/s	~935 ms ~940 ms ~1.0s
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	95 tok/s 79 tok/s 56 tok/s	~435 ms ~437 ms ~442 ms
-	Phi-1	1.3B	INT4 Q8 FP16	95 tok/s 79 tok/s 56 tok/s	~435 ms ~437 ms ~442 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	95 tok/s 79 tok/s 56 tok/s	~435 ms ~437 ms ~442 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	92 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	Falcon3-3B	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	LFM2-VL 3B	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	Liquid D1 3B	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	Liquid D1-VL 3B	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	Liquid LFM2-VL 3B D1	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	Qwen2.5-3B	3B	INT4 Q8 FP16	91 tok/s 61 tok/s 33 tok/s	~935 ms ~941 ms ~1.0s
-	Mellum 2.1 12B A2.5B Thinking	12B (2.5B active)	INT4	83 tok/s	~1.1s
-	FLUX.2 Klein 4B	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
-	FrogNano 4B	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
-	Mage-VL Cowork	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
-	Spark X2.5 4B	4B	INT4 Q8	75 tok/s 48 tok/s	~1.2s ~1.2s
-	EmbeddingGemma 2	2B	INT4 Q8 FP16	72 tok/s 58 tok/s 38 tok/s	~657 ms ~660 ms ~725 ms
-	CogVideoX 1.5 5B	5B	INT4 Q8	64 tok/s 38 tok/s	~1.5s ~1.6s
-	CogVideoX 5B	5B	INT4 Q8	64 tok/s 38 tok/s	~1.5s ~1.6s
-	Wan 2.2 TI2V 5B	5B	INT4 Q8	64 tok/s 38 tok/s	~1.5s ~1.6s
-	Phi-2	2.7B	INT4 Q8 FP16	63 tok/s 49 tok/s 29 tok/s	~856 ms ~860 ms ~1.0s
-	MaLLaM-3B	3B	INT4 Q8	60 tok/s 46 tok/s	~941 ms ~946 ms
-	SmolLM3 3B	3B	INT4 Q8	60 tok/s 46 tok/s	~941 ms ~946 ms
-	Falcon-7B	7B	INT4 Q8	52 tok/s 28 tok/s	~2.1s ~2.5s
-	AuraFlow v0.2	6.8B	INT4 Q8	51 tok/s 28 tok/s	~2.0s ~2.4s
-	AuraFlow v0.3	6.8B	INT4 Q8	51 tok/s 28 tok/s	~2.0s ~2.4s
-	Falcon3-7B	7B	INT4	49 tok/s	~2.1s
-	Hy-MT2 7B	7B	INT4	49 tok/s	~2.1s
-	Mistral-7B-v0.1	7.3B	INT4	48 tok/s	~2.2s
-	Youtu Parsing Omni	8B	INT4	46 tok/s	~2.4s
-	Stable Diffusion 3.5 Large	8.1B	INT4	44 tok/s	~2.4s
-	Stable Diffusion 3.5 Large Turbo	8.1B	INT4	44 tok/s	~2.4s
-	FLUX.2 Klein 9B	9B	INT4	40 tok/s	~2.7s
-	LensVLM 9B	9B	INT4	40 tok/s	~2.7s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4	40 tok/s	~2.7s
-	ChatGLM3-6B-32K	6B	INT4	37 tok/s	~2.0s
-	Yi-6B	6B	INT4	37 tok/s	~2.0s
-	Falcon3-10B	10B	INT4	35 tok/s	~3.2s
-	Mochi 1 Preview	10B	INT4	35 tok/s	~3.2s
-	Falcon2-11B	11B	INT4	34 tok/s	~3.5s
-	Hunyuan Lite	7B	INT4	33 tok/s	~2.3s
-	OLMo 3 7B Base	7B	INT4	33 tok/s	~2.3s
-	SEA-LION-7B	7.1B	INT4	33 tok/s	~2.3s
-	SEA-LION-7B-Instruct	7.1B	INT4	33 tok/s	~2.3s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4	29 tok/s	~2.8s
-	Typhoon-2-8B	8B	INT4	29 tok/s	~2.8s
-	FLUX.1 Canny Dev	11.9B	INT4	28 tok/s	~4.1s
-	FLUX.1 Depth Dev	11.9B	INT4	28 tok/s	~4.1s
-	FLUX.1 Dev	11.9B	INT4	28 tok/s	~4.1s
-	FLUX.1 Fill Dev	11.9B	INT4	28 tok/s	~4.1s
-	FLUX.1 Schnell	11.9B	INT4	28 tok/s	~4.1s
-	HunyuanVideo 1.5 720P T2V	13B	INT4	26 tok/s	~4.5s

AMD RX 7600 (8GB)

VRAM Scaling by Context Length

Best Models to Run on AMD RX 7600 (8GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions