What are the best models to run on AMD RX 7900 GRE (16GB)?

Based on model rankings and 16 GB usable VRAM: Top Overall: Qwen 3.8 27B (INT4) ~26 tok/s. Top Coding: Qwen3.5-27B (INT4) ~26 tok/s. Optimal Daily Driver: Phi-4 Reasoning Plus (INT4) ~38 tok/s. Maximum Model Size: ERNIE-4.5-VL-28B-A3B (INT4) ~120 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 7900 GRE (16GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#33	Qwen 3.8 27B	27B	INT4	26 tok/s	~3.8s
#83	Qwen3.5-27B	27B	INT4	26 tok/s	~3.8s
#100	Phi-4 Reasoning Plus	14B	INT4	38 tok/s	~1.7s
#101	MiMo V2 Flash	15B (309B active)	INT4	3 tok/s	~38.0s
#108	Gemma 4 26B A4B	25.2B (3.8B active)	INT4	114 tok/s	~628 ms
#109	Gemma 4 12B	11.95B	INT4 Q8	44 tok/s 27 tok/s	~1.5s ~1.7s
#113	Qwen3.5-9B	9B	INT4 Q8	75 tok/s 46 tok/s	~1.1s ~1.1s
#118	K2 Horizon 7B	7B	INT4 Q8 FP16	90 tok/s 57 tok/s 28 tok/s	~858 ms ~865 ms ~1.0s
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	25 tok/s 15 tok/s 8 tok/s	~3.4s ~3.5s ~3.5s
#132	Qwen3.5-4B	4B	INT4 Q8 FP16	132 tok/s 89 tok/s 51 tok/s	~506 ms ~509 ms ~517 ms
#136	GPT-OSS 20B	21B (3.6B active)	INT4	54 tok/s	~644 ms
#146	Granite 4.2 8B	8B	INT4 Q8	82 tok/s 51 tok/s	~976 ms ~983 ms
#150	Gemma 4 E4B	8B	INT4 Q8	82 tok/s 51 tok/s	~976 ms ~983 ms
#158	GLM-4V	9B	INT4 Q8	55 tok/s 36 tok/s	~1.1s ~1.2s
#159	Gemma 3 27B	27B	INT4	26 tok/s	~3.8s
#160	Phi-4	14B	INT4 Q8	52 tok/s 28 tok/s	~1.7s ~2.0s
#169	OLMo 3 7B Think	7B	INT4 Q8	66 tok/s 47 tok/s	~863 ms ~869 ms
#172	DeepSeek-R1 14B	14B	INT4 Q8	54 tok/s 28 tok/s	~1.7s ~2.0s
#173	Ministral 3 14B	14B	INT4	38 tok/s	~1.7s
#177	Gemma 3 12B	12B	INT4 Q8	59 tok/s 34 tok/s	~1.5s ~1.6s
#181	Mistral-Small-2501	24B	INT4	31 tok/s	~3.1s
#182	Ministral 3 8B	8B	INT4 Q8	60 tok/s 42 tok/s	~980 ms ~987 ms
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	189 tok/s 140 tok/s 89 tok/s	~271 ms ~273 ms ~277 ms
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	189 tok/s 140 tok/s 89 tok/s	~271 ms ~273 ms ~277 ms
#185	Qwen2.5-14B	14B	INT4 Q8	52 tok/s 28 tok/s	~1.7s ~2.0s
#186	Gemma 2 27B	27B	INT4	26 tok/s	~3.8s
#188	Gemma 4 E2B	5.1B	INT4 Q8 FP16	113 tok/s 74 tok/s 42 tok/s	~635 ms ~640 ms ~650 ms
#189	Qwen3-4B	4B	INT4 Q8 FP16	132 tok/s 89 tok/s 51 tok/s	~506 ms ~509 ms ~517 ms
#190	DeepSeek-R1 8B	8B	INT4 Q8	84 tok/s 52 tok/s	~976 ms ~983 ms
#193	Ministral 3 3B	3B	INT4 Q8 FP16	107 tok/s 84 tok/s 56 tok/s	~391 ms ~393 ms ~399 ms
#195	Qwen3-14B	14B	INT4 Q8	52 tok/s 28 tok/s	~1.7s ~2.0s
#197	Gemma 2 9B	9B	INT4 Q8	75 tok/s 46 tok/s	~1.1s ~1.1s
#198	Qwen2.5-7B	7B	INT4 Q8 FP16	90 tok/s 57 tok/s 28 tok/s	~858 ms ~865 ms ~1.0s
#199	Qwen3-8B	8B	INT4 Q8	82 tok/s 51 tok/s	~976 ms ~983 ms
#200	Gemma 3 4B	4B	INT4 Q8 FP16	132 tok/s 89 tok/s 51 tok/s	~506 ms ~509 ms ~517 ms
#201	DeepSeek-R1 7B	7B	INT4 Q8 FP16	93 tok/s 58 tok/s 30 tok/s	~858 ms ~865 ms ~954 ms
#202	Llama 3.1 8B	8B	INT4 Q8	82 tok/s 51 tok/s	~976 ms ~983 ms
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	203 tok/s 154 tok/s 101 tok/s	~233 ms ~234 ms ~238 ms
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	303 tok/s 283 tok/s 245 tok/s	~42 ms ~42 ms ~43 ms
#208	Llama 3 8B	8B	INT4 Q8	82 tok/s 51 tok/s	~976 ms ~983 ms
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	288 tok/s 249 tok/s 191 tok/s	~86 ms ~87 ms ~88 ms
#210	Phi-4-Mini	3.8B	INT4 Q8 FP16	136 tok/s 92 tok/s 54 tok/s	~482 ms ~486 ms ~493 ms
#212	Ministral-8B-2410	8B	INT4 Q8	82 tok/s 51 tok/s	~976 ms ~983 ms
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	215 tok/s 166 tok/s 110 tok/s	~207 ms ~208 ms ~211 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	155 tok/s 109 tok/s 65 tok/s	~388 ms ~391 ms ~397 ms
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	267 tok/s 223 tok/s 164 tok/s	~114 ms ~115 ms ~116 ms
#216	Phi-3-mini	3.8B	INT4 Q8 FP16	136 tok/s 92 tok/s 54 tok/s	~482 ms ~486 ms ~493 ms
#217	Gemma 3n E2B IT	6B (2B active)	INT4 Q8 FP16	107 tok/s 92 tok/s 64 tok/s	~293 ms ~294 ms ~324 ms
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8 FP16	88 tok/s 55 tok/s 27 tok/s	~893 ms ~900 ms ~1.1s
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	248 tok/s 202 tok/s 143 tok/s	~143 ms ~144 ms ~146 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	189 tok/s 140 tok/s 89 tok/s	~271 ms ~273 ms ~277 ms
#221	Phi-3-medium	14B	INT4 Q8	52 tok/s 28 tok/s	~1.7s ~2.0s
#222	OLMo 3 7B Instruct	7B	INT4 Q8	66 tok/s 47 tok/s	~863 ms ~869 ms
#224	Qwen2-7B	7B	INT4 Q8 FP16	90 tok/s 57 tok/s 28 tok/s	~858 ms ~865 ms ~1.0s
#225	Phi-3-small	7B	INT4 Q8 FP16	90 tok/s 57 tok/s 28 tok/s	~858 ms ~865 ms ~1.0s
#226	Gemma 3 1B	1B	INT4 Q8 FP16	248 tok/s 202 tok/s 143 tok/s	~143 ms ~144 ms ~146 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8 FP16	88 tok/s 55 tok/s 27 tok/s	~893 ms ~900 ms ~1.1s
#229	Gemma 1 7B	7B	INT4 Q8	66 tok/s 47 tok/s	~863 ms ~869 ms
#230	Gemma 1 2B	2B	INT4 Q8 FP16	200 tok/s 146 tok/s 91 tok/s	~271 ms ~273 ms ~277 ms
#231	ChatGLM3-6B	6B	INT4 Q8 FP16	73 tok/s 53 tok/s 30 tok/s	~745 ms ~750 ms ~827 ms
#232	ChatGLM2-6B	6B	INT4 Q8 FP16	73 tok/s 53 tok/s 30 tok/s	~745 ms ~750 ms ~827 ms
#233	ChatGLM-6B	6B	INT4 Q8 FP16	73 tok/s 53 tok/s 30 tok/s	~745 ms ~750 ms ~827 ms
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	300 tok/s 263 tok/s 208 tok/s	~72 ms ~73 ms ~74 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	300 tok/s 263 tok/s 208 tok/s	~72 ms ~73 ms ~74 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	295 tok/s 275 tok/s 235 tok/s	~46 ms ~47 ms ~47 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	295 tok/s 275 tok/s 235 tok/s	~46 ms ~47 ms ~47 ms
-	Falcon-1B	1B	INT4 Q8 FP16	259 tok/s 209 tok/s 146 tok/s	~143 ms ~144 ms ~146 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	248 tok/s 202 tok/s 143 tok/s	~143 ms ~144 ms ~146 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	215 tok/s 166 tok/s 110 tok/s	~207 ms ~208 ms ~211 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	215 tok/s 166 tok/s 110 tok/s	~207 ms ~208 ms ~211 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	203 tok/s 154 tok/s 101 tok/s	~233 ms ~234 ms ~238 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	199 tok/s 150 tok/s 97 tok/s	~245 ms ~247 ms ~250 ms
-	Falcon-3B	3B	INT4 Q8 FP16	163 tok/s 112 tok/s 66 tok/s	~388 ms ~390 ms ~396 ms
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	161 tok/s 138 tok/s 101 tok/s	~183 ms ~184 ms ~186 ms
-	Phi-1	1.3B	INT4 Q8 FP16	161 tok/s 138 tok/s 101 tok/s	~183 ms ~184 ms ~186 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	161 tok/s 138 tok/s 101 tok/s	~183 ms ~184 ms ~186 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	157 tok/s 109 tok/s 66 tok/s	~388 ms ~391 ms ~397 ms
-	Falcon3-3B	3B	INT4 Q8 FP16	155 tok/s 109 tok/s 65 tok/s	~388 ms ~391 ms ~397 ms
-	Qwen2.5-3B	3B	INT4 Q8 FP16	155 tok/s 109 tok/s 65 tok/s	~388 ms ~391 ms ~397 ms
-	Spark X2.5 4B	4B	INT4 Q8 FP16	132 tok/s 89 tok/s 51 tok/s	~506 ms ~509 ms ~517 ms
-	ERNIE-4.5-21B-A3B	21B (3B active)	INT4	131 tok/s	~513 ms
-	ERNIE-4.5-21B-A3B-Base	21B (3B active)	INT4	131 tok/s	~513 ms
-	ERNIE-4.5-VL-28B-A3B	28B (3B active)	INT4	120 tok/s	~583 ms
-	ERNIE-4.5-VL-28B-A3B-Base	28B (3B active)	INT4	120 tok/s	~583 ms
-	Phi-2	2.7B	INT4 Q8 FP16	113 tok/s 90 tok/s 60 tok/s	~356 ms ~358 ms ~364 ms
-	MaLLaM-3B	3B	INT4 Q8 FP16	107 tok/s 84 tok/s 56 tok/s	~391 ms ~393 ms ~399 ms
-	SmolLM3 3B	3B	INT4 Q8 FP16	107 tok/s 84 tok/s 56 tok/s	~391 ms ~393 ms ~399 ms
-	Falcon-7B	7B	INT4 Q8 FP16	94 tok/s 58 tok/s 30 tok/s	~858 ms ~864 ms ~953 ms
-	Falcon3-7B	7B	INT4 Q8 FP16	90 tok/s 57 tok/s 28 tok/s	~858 ms ~865 ms ~1.0s
-	Hy-MT2 7B	7B	INT4 Q8 FP16	90 tok/s 57 tok/s 28 tok/s	~858 ms ~865 ms ~1.0s
-	Mistral-7B-v0.1	7.3B	INT4 Q8 FP16	88 tok/s 55 tok/s 27 tok/s	~893 ms ~900 ms ~1.1s
-	LensVLM 9B	9B	INT4 Q8	75 tok/s 46 tok/s	~1.1s ~1.1s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8	75 tok/s 46 tok/s	~1.1s ~1.1s
-	ChatGLM3-6B-32K	6B	INT4 Q8 FP16	73 tok/s 53 tok/s 30 tok/s	~745 ms ~750 ms ~827 ms
-	Yi-6B	6B	INT4 Q8 FP16	73 tok/s 53 tok/s 30 tok/s	~745 ms ~750 ms ~827 ms
-	Falcon3-10B	10B	INT4 Q8	69 tok/s 42 tok/s	~1.2s ~1.2s
-	Falcon2-11B	11B	INT4 Q8	66 tok/s 39 tok/s	~1.3s ~1.3s
-	Hunyuan Lite	7B	INT4 Q8	66 tok/s 47 tok/s	~863 ms ~869 ms
-	OLMo 3 7B Base	7B	INT4 Q8	66 tok/s 47 tok/s	~863 ms ~869 ms
-	Kimi-VL-A3B-Instruct	16B (3B active)	INT4	65 tok/s	~500 ms
-	SEA-LION-7B	7.1B	INT4 Q8	65 tok/s 46 tok/s	~874 ms ~880 ms
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8	65 tok/s 46 tok/s	~874 ms ~880 ms
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8	60 tok/s 42 tok/s	~980 ms ~987 ms
-	Typhoon-2-8B	8B	INT4 Q8	60 tok/s 42 tok/s	~980 ms ~987 ms
-	GLM-4-9B	9B	INT4 Q8	55 tok/s 36 tok/s	~1.1s ~1.2s
-	GLM-4-9B-Chat	9B	INT4 Q8	55 tok/s 36 tok/s	~1.1s ~1.2s
-	GLM-4-9B-Chat-1M	9B	INT4 Q8	55 tok/s 36 tok/s	~1.1s ~1.2s
-	Yi-9B	9B	INT4 Q8	55 tok/s 36 tok/s	~1.1s ~1.2s
-	Sahabat-AI-Gemma2-9B	9.2B	INT4 Q8	54 tok/s 36 tok/s	~1.1s ~1.2s
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4 Q8	54 tok/s 36 tok/s	~1.1s ~1.2s
-	GreenMind-14B-R1	14B	INT4	38 tok/s	~1.7s

AMD RX 7900 GRE (16GB)

VRAM Scaling by Context Length

Best Models to Run on AMD RX 7900 GRE (16GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions