What are the best models to run on AMD RX 6700 XT (12GB)?

Based on model rankings and 12 GB usable VRAM: Top Overall: Phi-4 Reasoning Plus (INT4) ~24 tok/s. Top Coding: K2 Horizon 7B (INT4) ~66 tok/s. Optimal Daily Driver: Gemma 4 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 6700 XT (12GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#100	Phi-4 Reasoning Plus	14B	INT4	24 tok/s	~7.0s
#109	Gemma 4 12B	11.95B	INT4	29 tok/s	~5.5s
#113	Qwen3.5-9B	9B	INT4 Q8	53 tok/s 30 tok/s	~3.8s ~4.2s
#118	K2 Horizon 7B	7B	INT4 Q8	66 tok/s 40 tok/s	~3.0s ~3.0s
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	17 tok/s 10 tok/s 5 tok/s	~12.0s ~12.1s ~13.2s
#132	Qwen3.5-4B	4B	INT4 Q8 FP16	100 tok/s 64 tok/s 36 tok/s	~1.8s ~1.8s ~1.8s
#146	Granite 4.2 8B	8B	INT4 Q8	59 tok/s 34 tok/s	~3.4s ~3.7s
#150	Gemma 4 E4B	8B	INT4 Q8	59 tok/s 34 tok/s	~3.4s ~3.7s
#158	GLM-4V	9B	INT4	38 tok/s	~3.8s
#160	Phi-4	14B	INT4	36 tok/s	~5.9s
#169	OLMo 3 7B Think	7B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
#172	DeepSeek-R1 14B	14B	INT4	38 tok/s	~5.9s
#173	Ministral 3 14B	14B	INT4	24 tok/s	~7.0s
#177	Gemma 3 12B	12B	INT4	42 tok/s	~5.1s
#182	Ministral 3 8B	8B	INT4 Q8	42 tok/s 26 tok/s	~3.4s ~4.0s
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	152 tok/s 107 tok/s 65 tok/s	~932 ms ~935 ms ~941 ms
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	152 tok/s 107 tok/s 65 tok/s	~932 ms ~935 ms ~941 ms
#185	Qwen2.5-14B	14B	INT4	36 tok/s	~5.9s
#188	Gemma 4 E2B	5.1B	INT4 Q8 FP16	84 tok/s 53 tok/s 26 tok/s	~2.2s ~2.2s ~2.6s
#189	Qwen3-4B	4B	INT4 Q8 FP16	100 tok/s 64 tok/s 36 tok/s	~1.8s ~1.8s ~1.8s
#190	DeepSeek-R1 8B	8B	INT4 Q8	60 tok/s 36 tok/s	~3.4s ~3.4s
#193	Ministral 3 3B	3B	INT4 Q8 FP16	79 tok/s 61 tok/s 39 tok/s	~1.3s ~1.4s ~1.4s
#195	Qwen3-14B	14B	INT4	36 tok/s	~5.9s
#197	Gemma 2 9B	9B	INT4 Q8	53 tok/s 30 tok/s	~3.8s ~4.2s
#198	Qwen2.5-7B	7B	INT4 Q8	66 tok/s 40 tok/s	~3.0s ~3.0s
#199	Qwen3-8B	8B	INT4 Q8	59 tok/s 34 tok/s	~3.4s ~3.7s
#200	Gemma 3 4B	4B	INT4 Q8 FP16	100 tok/s 64 tok/s 36 tok/s	~1.8s ~1.8s ~1.8s
#201	DeepSeek-R1 7B	7B	INT4 Q8	67 tok/s 41 tok/s	~3.0s ~3.0s
#202	Llama 3.1 8B	8B	INT4 Q8	59 tok/s 34 tok/s	~3.4s ~3.7s
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	166 tok/s 120 tok/s 74 tok/s	~798 ms ~800 ms ~805 ms
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	277 tok/s 253 tok/s 210 tok/s	~129 ms ~129 ms ~130 ms
#208	Llama 3 8B	8B	INT4 Q8	59 tok/s 34 tok/s	~3.4s ~3.7s
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	259 tok/s 214 tok/s 153 tok/s	~283 ms ~283 ms ~285 ms
#210	Phi-4-Mini	3.8B	INT4 Q8 FP16	103 tok/s 67 tok/s 38 tok/s	~1.7s ~1.7s ~1.7s
#212	Ministral-8B-2410	8B	INT4 Q8	59 tok/s 34 tok/s	~3.4s ~3.7s
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	177 tok/s 130 tok/s 81 tok/s	~706 ms ~708 ms ~713 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	120 tok/s 80 tok/s 46 tok/s	~1.3s ~1.3s ~1.4s
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	234 tok/s 186 tok/s 128 tok/s	~381 ms ~382 ms ~384 ms
#216	Phi-3-mini	3.8B	INT4 Q8 FP16	103 tok/s 67 tok/s 38 tok/s	~1.7s ~1.7s ~1.7s
#217	Gemma 3n E2B IT	6B (2B active)	INT4 Q8	79 tok/s 67 tok/s	~1.0s ~1.0s
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8	63 tok/s 39 tok/s	~3.1s ~3.1s
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	213 tok/s 165 tok/s 109 tok/s	~483 ms ~485 ms ~488 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	152 tok/s 107 tok/s 65 tok/s	~932 ms ~935 ms ~941 ms
#221	Phi-3-medium	14B	INT4	36 tok/s	~5.9s
#222	OLMo 3 7B Instruct	7B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
#224	Qwen2-7B	7B	INT4 Q8	66 tok/s 40 tok/s	~3.0s ~3.0s
#225	Phi-3-small	7B	INT4 Q8	66 tok/s 40 tok/s	~3.0s ~3.0s
#226	Gemma 3 1B	1B	INT4 Q8 FP16	213 tok/s 165 tok/s 109 tok/s	~483 ms ~485 ms ~488 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8	63 tok/s 39 tok/s	~3.1s ~3.1s
#229	Gemma 1 7B	7B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
#230	Gemma 1 2B	2B	INT4 Q8 FP16	162 tok/s 112 tok/s 66 tok/s	~932 ms ~935 ms ~941 ms
#231	ChatGLM3-6B	6B	INT4 Q8	52 tok/s 37 tok/s	~2.6s ~2.6s
#232	ChatGLM2-6B	6B	INT4 Q8	52 tok/s 37 tok/s	~2.6s ~2.6s
#233	ChatGLM-6B	6B	INT4 Q8	52 tok/s 37 tok/s	~2.6s ~2.6s
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	274 tok/s 230 tok/s 170 tok/s	~235 ms ~235 ms ~237 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	274 tok/s 230 tok/s 170 tok/s	~235 ms ~235 ms ~237 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	268 tok/s 243 tok/s 199 tok/s	~143 ms ~143 ms ~144 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	268 tok/s 243 tok/s 199 tok/s	~143 ms ~143 ms ~144 ms
-	Falcon-1B	1B	INT4 Q8 FP16	225 tok/s 171 tok/s 112 tok/s	~483 ms ~484 ms ~488 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	213 tok/s 165 tok/s 109 tok/s	~483 ms ~485 ms ~488 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	177 tok/s 130 tok/s 81 tok/s	~706 ms ~708 ms ~713 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	177 tok/s 130 tok/s 81 tok/s	~706 ms ~708 ms ~713 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	166 tok/s 120 tok/s 74 tok/s	~798 ms ~800 ms ~805 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	161 tok/s 115 tok/s 70 tok/s	~840 ms ~842 ms ~848 ms
-	Falcon-3B	3B	INT4 Q8 FP16	128 tok/s 83 tok/s 47 tok/s	~1.3s ~1.3s ~1.4s
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	125 tok/s 105 tok/s 74 tok/s	~618 ms ~620 ms ~624 ms
-	Phi-1	1.3B	INT4 Q8 FP16	125 tok/s 105 tok/s 74 tok/s	~618 ms ~620 ms ~624 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	125 tok/s 105 tok/s 74 tok/s	~618 ms ~620 ms ~624 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	122 tok/s 81 tok/s 46 tok/s	~1.3s ~1.3s ~1.4s
-	Falcon3-3B	3B	INT4 Q8 FP16	120 tok/s 80 tok/s 46 tok/s	~1.3s ~1.3s ~1.4s
-	Qwen2.5-3B	3B	INT4 Q8 FP16	120 tok/s 80 tok/s 46 tok/s	~1.3s ~1.3s ~1.4s
-	Spark X2.5 4B	4B	INT4 Q8 FP16	100 tok/s 64 tok/s 36 tok/s	~1.8s ~1.8s ~1.8s
-	Phi-2	2.7B	INT4 Q8 FP16	83 tok/s 65 tok/s 42 tok/s	~1.2s ~1.2s ~1.2s
-	MaLLaM-3B	3B	INT4 Q8 FP16	79 tok/s 61 tok/s 39 tok/s	~1.3s ~1.4s ~1.4s
-	SmolLM3 3B	3B	INT4 Q8 FP16	79 tok/s 61 tok/s 39 tok/s	~1.3s ~1.4s ~1.4s
-	Falcon-7B	7B	INT4 Q8	69 tok/s 41 tok/s	~3.0s ~3.0s
-	Falcon3-7B	7B	INT4 Q8	66 tok/s 40 tok/s	~3.0s ~3.0s
-	Hy-MT2 7B	7B	INT4 Q8	66 tok/s 40 tok/s	~3.0s ~3.0s
-	Mistral-7B-v0.1	7.3B	INT4 Q8	63 tok/s 39 tok/s	~3.1s ~3.1s
-	LensVLM 9B	9B	INT4 Q8	53 tok/s 30 tok/s	~3.8s ~4.2s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8	53 tok/s 30 tok/s	~3.8s ~4.2s
-	ChatGLM3-6B-32K	6B	INT4 Q8	52 tok/s 37 tok/s	~2.6s ~2.6s
-	Yi-6B	6B	INT4 Q8	52 tok/s 37 tok/s	~2.6s ~2.6s
-	Falcon3-10B	10B	INT4 Q8	49 tok/s 26 tok/s	~4.2s ~5.0s
-	Falcon2-11B	11B	INT4 Q8	47 tok/s 25 tok/s	~4.6s ~5.5s
-	Hunyuan Lite	7B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
-	OLMo 3 7B Base	7B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
-	SEA-LION-7B	7.1B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8	46 tok/s 31 tok/s	~3.0s ~3.3s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8	42 tok/s 26 tok/s	~3.4s ~4.0s
-	Typhoon-2-8B	8B	INT4 Q8	42 tok/s 26 tok/s	~3.4s ~4.0s
-	GLM-4-9B	9B	INT4	38 tok/s	~3.8s
-	GLM-4-9B-Chat	9B	INT4	38 tok/s	~3.8s
-	GLM-4-9B-Chat-1M	9B	INT4	38 tok/s	~3.8s
-	Sahabat-AI-Gemma2-9B	9.2B	INT4	38 tok/s	~3.9s
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4	38 tok/s	~3.9s
-	Yi-9B	9B	INT4	38 tok/s	~3.8s
-	GreenMind-14B-R1	14B	INT4	24 tok/s	~7.0s

AMD RX 6700 XT (12GB)

VRAM Scaling by Context Length

Best Models to Run on AMD RX 6700 XT (12GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions