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) ~42 tok/s. Top Coding: K2 Horizon 7B (INT4) ~51 tok/s. Optimal Daily Driver: Granite 4.2 8B (INT4) ~46 tok/s. Maximum Model Size: Gemma 3 12B (INT4) ~29 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 AMD RX 7600 (8GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#113	Qwen3.5-9B	9B	INT4	42 tok/s	~2.3s
#118	K2 Horizon 7B	7B	INT4	51 tok/s	~1.8s
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8	13 tok/s 7 tok/s	~7.3s ~8.0s
#132	Qwen3.5-4B	4B	INT4 Q8	80 tok/s 50 tok/s	~1.1s ~1.1s
#146	Granite 4.2 8B	8B	INT4	46 tok/s	~2.1s
#150	Gemma 4 E4B	8B	INT4	46 tok/s	~2.1s
#169	OLMo 3 7B Think	7B	INT4	34 tok/s	~2.0s
#177	Gemma 3 12B	12B	INT4	29 tok/s	~3.6s
#182	Ministral 3 8B	8B	INT4	29 tok/s	~2.4s
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	127 tok/s 87 tok/s 51 tok/s	~570 ms ~574 ms ~582 ms
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	127 tok/s 87 tok/s 51 tok/s	~570 ms ~574 ms ~582 ms
#188	Gemma 4 E2B	5.1B	INT4 Q8	67 tok/s 39 tok/s	~1.3s ~1.5s
#189	Qwen3-4B	4B	INT4 Q8	80 tok/s 50 tok/s	~1.1s ~1.1s
#190	DeepSeek-R1 8B	8B	INT4	47 tok/s	~2.1s
#193	Ministral 3 3B	3B	INT4 Q8	63 tok/s 48 tok/s	~825 ms ~830 ms
#197	Gemma 2 9B	9B	INT4	42 tok/s	~2.3s
#198	Qwen2.5-7B	7B	INT4	51 tok/s	~1.8s
#199	Qwen3-8B	8B	INT4	46 tok/s	~2.1s
#200	Gemma 3 4B	4B	INT4 Q8	80 tok/s 50 tok/s	~1.1s ~1.1s
#201	DeepSeek-R1 7B	7B	INT4 Q8	53 tok/s 28 tok/s	~1.8s ~2.2s
#202	Llama 3.1 8B	8B	INT4	46 tok/s	~2.1s
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	140 tok/s 98 tok/s 58 tok/s	~488 ms ~491 ms ~498 ms
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	256 tok/s 229 tok/s 183 tok/s	~81 ms ~82 ms ~83 ms
#208	Llama 3 8B	8B	INT4	46 tok/s	~2.1s
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	236 tok/s 187 tok/s 128 tok/s	~175 ms ~176 ms ~178 ms
#210	Phi-4-Mini	3.8B	INT4 Q8	83 tok/s 53 tok/s	~1.0s ~1.0s
#212	Ministral-8B-2410	8B	INT4	46 tok/s	~2.1s
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	151 tok/s 107 tok/s 65 tok/s	~432 ms ~435 ms ~441 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	98 tok/s 64 tok/s 34 tok/s	~819 ms ~825 ms ~908 ms
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	209 tok/s 160 tok/s 105 tok/s	~235 ms ~236 ms ~239 ms
#216	Phi-3-mini	3.8B	INT4 Q8	83 tok/s 53 tok/s	~1.0s ~1.0s
#217	Gemma 3n E2B IT	6B (2B active)	INT4	60 tok/s	~669 ms
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4	50 tok/s	~1.9s
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	186 tok/s 139 tok/s 89 tok/s	~297 ms ~299 ms ~303 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	127 tok/s 87 tok/s 51 tok/s	~570 ms ~574 ms ~582 ms
#222	OLMo 3 7B Instruct	7B	INT4	34 tok/s	~2.0s
#224	Qwen2-7B	7B	INT4	51 tok/s	~1.8s
#225	Phi-3-small	7B	INT4	51 tok/s	~1.8s
#226	Gemma 3 1B	1B	INT4 Q8 FP16	186 tok/s 139 tok/s 89 tok/s	~297 ms ~299 ms ~303 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4	50 tok/s	~1.9s
#229	Gemma 1 7B	7B	INT4	34 tok/s	~2.0s
#230	Gemma 1 2B	2B	INT4 Q8 FP16	137 tok/s 91 tok/s 52 tok/s	~569 ms ~573 ms ~581 ms
#231	ChatGLM3-6B	6B	INT4	38 tok/s	~1.7s
#232	ChatGLM2-6B	6B	INT4	38 tok/s	~1.7s
#233	ChatGLM-6B	6B	INT4	38 tok/s	~1.7s
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	252 tok/s 204 tok/s 144 tok/s	~146 ms ~147 ms ~149 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	252 tok/s 204 tok/s 144 tok/s	~146 ms ~147 ms ~149 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	245 tok/s 218 tok/s 172 tok/s	~90 ms ~91 ms ~92 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	245 tok/s 218 tok/s 172 tok/s	~90 ms ~91 ms ~92 ms
-	Falcon-1B	1B	INT4 Q8 FP16	199 tok/s 145 tok/s 91 tok/s	~297 ms ~298 ms ~303 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	186 tok/s 139 tok/s 89 tok/s	~297 ms ~299 ms ~303 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	151 tok/s 107 tok/s 65 tok/s	~432 ms ~435 ms ~441 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	151 tok/s 107 tok/s 65 tok/s	~432 ms ~435 ms ~441 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	140 tok/s 98 tok/s 58 tok/s	~488 ms ~491 ms ~498 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	136 tok/s 94 tok/s 55 tok/s	~514 ms ~517 ms ~524 ms
-	Falcon-3B	3B	INT4 Q8 FP16	105 tok/s 66 tok/s 35 tok/s	~818 ms ~824 ms ~907 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	103 tok/s 84 tok/s 58 tok/s	~381 ms ~383 ms ~388 ms
-	Phi-1	1.3B	INT4 Q8 FP16	103 tok/s 84 tok/s 58 tok/s	~381 ms ~383 ms ~388 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	103 tok/s 84 tok/s 58 tok/s	~381 ms ~383 ms ~388 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	100 tok/s 64 tok/s 34 tok/s	~819 ms ~824 ms ~908 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	98 tok/s 64 tok/s 34 tok/s	~819 ms ~825 ms ~908 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	98 tok/s 64 tok/s 34 tok/s	~819 ms ~825 ms ~908 ms
-	Spark X2.5 4B	4B	INT4 Q8	80 tok/s 50 tok/s	~1.1s ~1.1s
-	Phi-2	2.7B	INT4 Q8 FP16	66 tok/s 51 tok/s 29 tok/s	~750 ms ~754 ms ~897 ms
-	MaLLaM-3B	3B	INT4 Q8	63 tok/s 48 tok/s	~825 ms ~830 ms
-	SmolLM3 3B	3B	INT4 Q8	63 tok/s 48 tok/s	~825 ms ~830 ms
-	Falcon-7B	7B	INT4 Q8	54 tok/s 29 tok/s	~1.8s ~2.2s
-	Falcon3-7B	7B	INT4	51 tok/s	~1.8s
-	Hy-MT2 7B	7B	INT4	51 tok/s	~1.8s
-	Mistral-7B-v0.1	7.3B	INT4	50 tok/s	~1.9s
-	LensVLM 9B	9B	INT4	42 tok/s	~2.3s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4	42 tok/s	~2.3s
-	ChatGLM3-6B-32K	6B	INT4	38 tok/s	~1.7s
-	Yi-6B	6B	INT4	38 tok/s	~1.7s
-	Falcon3-10B	10B	INT4	36 tok/s	~2.8s
-	Falcon2-11B	11B	INT4	35 tok/s	~3.1s
-	Hunyuan Lite	7B	INT4	34 tok/s	~2.0s
-	OLMo 3 7B Base	7B	INT4	34 tok/s	~2.0s
-	SEA-LION-7B	7.1B	INT4	34 tok/s	~2.0s
-	SEA-LION-7B-Instruct	7.1B	INT4	34 tok/s	~2.0s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4	29 tok/s	~2.4s
-	Typhoon-2-8B	8B	INT4	29 tok/s	~2.4s

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