What are the best models to run on AMD RX 6700 (10GB)?

Based on model rankings and 10 GB usable VRAM: Top Overall: Qwen3.5-9B (INT4) ~46 tok/s. Top Coding: K2 Horizon 7B (INT4) ~56 tok/s. Optimal Daily Driver: Granite 4.2 8B (INT4) ~50 tok/s. Maximum Model Size: DeepSeek-R1 14B (INT4) ~30 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 (10GB) (October 2026)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#113	Qwen3.5-9B	9B	INT4	46 tok/s	~4.4s
#118	K2 Horizon 7B	7B	INT4 Q8	56 tok/s 32 tok/s	~3.4s ~3.8s
#128	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	15 tok/s 8 tok/s 4 tok/s	~13.9s ~13.9s ~16.5s
#132	Qwen3.5-4B	4B	INT4 Q8 FP16	87 tok/s 55 tok/s 29 tok/s	~2.0s ~2.0s ~2.2s
#146	Granite 4.2 8B	8B	INT4 Q8	50 tok/s 27 tok/s	~3.9s ~4.6s
#150	Gemma 4 E4B	8B	INT4 Q8	50 tok/s 27 tok/s	~3.9s ~4.6s
#158	GLM-4V	9B	INT4	31 tok/s	~4.8s
#160	Phi-4	14B	INT4	29 tok/s	~7.4s
#169	OLMo 3 7B Think	7B	INT4 Q8	40 tok/s 25 tok/s	~3.4s ~4.1s
#172	DeepSeek-R1 14B	14B	INT4	30 tok/s	~7.4s
#177	Gemma 3 12B	12B	INT4	35 tok/s	~5.8s
#182	Ministral 3 8B	8B	INT4	36 tok/s	~3.9s
#183	Qwen3.5-2B	2B	INT4 Q8 FP16	136 tok/s 94 tok/s 55 tok/s	~1.1s ~1.1s ~1.1s
#184	MiniCPM5-2B	2B	INT4 Q8 FP16	136 tok/s 94 tok/s 55 tok/s	~1.1s ~1.1s ~1.1s
#185	Qwen2.5-14B	14B	INT4	29 tok/s	~7.4s
#188	Gemma 4 E2B	5.1B	INT4 Q8	72 tok/s 45 tok/s	~2.5s ~2.5s
#189	Qwen3-4B	4B	INT4 Q8 FP16	87 tok/s 55 tok/s 29 tok/s	~2.0s ~2.0s ~2.2s
#190	DeepSeek-R1 8B	8B	INT4 Q8	52 tok/s 29 tok/s	~3.9s ~4.3s
#193	Ministral 3 3B	3B	INT4 Q8 FP16	69 tok/s 52 tok/s 31 tok/s	~1.5s ~1.6s ~1.7s
#195	Qwen3-14B	14B	INT4	29 tok/s	~7.4s
#197	Gemma 2 9B	9B	INT4	46 tok/s	~4.4s
#198	Qwen2.5-7B	7B	INT4 Q8	56 tok/s 32 tok/s	~3.4s ~3.8s
#199	Qwen3-8B	8B	INT4 Q8	50 tok/s 27 tok/s	~3.9s ~4.6s
#200	Gemma 3 4B	4B	INT4 Q8 FP16	87 tok/s 55 tok/s 29 tok/s	~2.0s ~2.0s ~2.2s
#201	DeepSeek-R1 7B	7B	INT4 Q8	58 tok/s 33 tok/s	~3.4s ~3.8s
#202	Llama 3.1 8B	8B	INT4 Q8	50 tok/s 27 tok/s	~3.9s ~4.6s
#204	Qwen3-1.7B	1.7B	INT4 Q8 FP16	150 tok/s 105 tok/s 64 tok/s	~917 ms ~920 ms ~926 ms
#206	Gemma 3 270M	0.27B	INT4 Q8 FP16	264 tok/s 238 tok/s 193 tok/s	~147 ms ~147 ms ~148 ms
#208	Llama 3 8B	8B	INT4 Q8	50 tok/s 27 tok/s	~3.9s ~4.6s
#209	Qwen3-0.6B	0.6B	INT4 Q8 FP16	245 tok/s 197 tok/s 137 tok/s	~324 ms ~325 ms ~327 ms
#210	Phi-4-Mini	3.8B	INT4 Q8 FP16	90 tok/s 57 tok/s 30 tok/s	~1.9s ~1.9s ~2.1s
#212	Ministral-8B-2410	8B	INT4 Q8	50 tok/s 27 tok/s	~3.9s ~4.6s
#213	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	161 tok/s 115 tok/s 70 tok/s	~811 ms ~814 ms ~819 ms
#214	Llama 3.2 3B	3B	INT4 Q8 FP16	106 tok/s 69 tok/s 39 tok/s	~1.5s ~1.5s ~1.6s
#215	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	218 tok/s 169 tok/s 113 tok/s	~437 ms ~438 ms ~441 ms
#216	Phi-3-mini	3.8B	INT4 Q8 FP16	90 tok/s 57 tok/s 30 tok/s	~1.9s ~1.9s ~2.1s
#217	Gemma 3n E2B IT	6B (2B active)	INT4 Q8	68 tok/s 55 tok/s	~1.2s ~1.3s
#218	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8	54 tok/s 31 tok/s	~3.6s ~3.9s
#219	Llama 3.2 1B	1B	INT4 Q8 FP16	196 tok/s 148 tok/s 96 tok/s	~555 ms ~557 ms ~560 ms
#220	Gemma 2 2B	2B	INT4 Q8 FP16	136 tok/s 94 tok/s 55 tok/s	~1.1s ~1.1s ~1.1s
#221	Phi-3-medium	14B	INT4	29 tok/s	~7.4s
#222	OLMo 3 7B Instruct	7B	INT4 Q8	40 tok/s 25 tok/s	~3.4s ~4.1s
#224	Qwen2-7B	7B	INT4 Q8	56 tok/s 32 tok/s	~3.4s ~3.8s
#225	Phi-3-small	7B	INT4 Q8	56 tok/s 32 tok/s	~3.4s ~3.8s
#226	Gemma 3 1B	1B	INT4 Q8 FP16	196 tok/s 148 tok/s 96 tok/s	~555 ms ~557 ms ~560 ms
#227	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8	54 tok/s 31 tok/s	~3.6s ~3.9s
#229	Gemma 1 7B	7B	INT4 Q8	40 tok/s 25 tok/s	~3.4s ~4.1s
#230	Gemma 1 2B	2B	INT4 Q8 FP16	146 tok/s 98 tok/s 57 tok/s	~1.1s ~1.1s ~1.1s
#231	ChatGLM3-6B	6B	INT4 Q8	44 tok/s 30 tok/s	~3.0s ~3.2s
#232	ChatGLM2-6B	6B	INT4 Q8	44 tok/s 30 tok/s	~3.0s ~3.2s
#233	ChatGLM-6B	6B	INT4 Q8	44 tok/s 30 tok/s	~3.0s ~3.2s
Unranked Models (Sorted by Speed)
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	260 tok/s 214 tok/s 154 tok/s	~269 ms ~270 ms ~272 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	260 tok/s 214 tok/s 154 tok/s	~269 ms ~270 ms ~272 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	254 tok/s 228 tok/s 182 tok/s	~163 ms ~164 ms ~165 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	254 tok/s 228 tok/s 182 tok/s	~163 ms ~164 ms ~165 ms
-	Falcon-1B	1B	INT4 Q8 FP16	209 tok/s 155 tok/s 99 tok/s	~555 ms ~556 ms ~560 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	196 tok/s 148 tok/s 96 tok/s	~555 ms ~557 ms ~560 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	161 tok/s 115 tok/s 70 tok/s	~811 ms ~814 ms ~819 ms
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	161 tok/s 115 tok/s 70 tok/s	~811 ms ~814 ms ~819 ms
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	150 tok/s 105 tok/s 64 tok/s	~917 ms ~920 ms ~926 ms
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	145 tok/s 101 tok/s 61 tok/s	~965 ms ~968 ms ~975 ms
-	Falcon-3B	3B	INT4 Q8 FP16	113 tok/s 72 tok/s 40 tok/s	~1.5s ~1.5s ~1.6s
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	111 tok/s 92 tok/s 64 tok/s	~710 ms ~712 ms ~717 ms
-	Phi-1	1.3B	INT4 Q8 FP16	111 tok/s 92 tok/s 64 tok/s	~710 ms ~712 ms ~717 ms
-	Phi-1.5	1.3B	INT4 Q8 FP16	111 tok/s 92 tok/s 64 tok/s	~710 ms ~712 ms ~717 ms
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	108 tok/s 70 tok/s 40 tok/s	~1.5s ~1.5s ~1.6s
-	Falcon3-3B	3B	INT4 Q8 FP16	106 tok/s 69 tok/s 39 tok/s	~1.5s ~1.5s ~1.6s
-	Qwen2.5-3B	3B	INT4 Q8 FP16	106 tok/s 69 tok/s 39 tok/s	~1.5s ~1.5s ~1.6s
-	Spark X2.5 4B	4B	INT4 Q8 FP16	87 tok/s 55 tok/s 29 tok/s	~2.0s ~2.0s ~2.2s
-	Phi-2	2.7B	INT4 Q8 FP16	72 tok/s 56 tok/s 36 tok/s	~1.4s ~1.4s ~1.4s
-	MaLLaM-3B	3B	INT4 Q8 FP16	69 tok/s 52 tok/s 31 tok/s	~1.5s ~1.6s ~1.7s
-	SmolLM3 3B	3B	INT4 Q8 FP16	69 tok/s 52 tok/s 31 tok/s	~1.5s ~1.6s ~1.7s
-	Falcon-7B	7B	INT4 Q8	59 tok/s 33 tok/s	~3.4s ~3.8s
-	Falcon3-7B	7B	INT4 Q8	56 tok/s 32 tok/s	~3.4s ~3.8s
-	Hy-MT2 7B	7B	INT4 Q8	56 tok/s 32 tok/s	~3.4s ~3.8s
-	Mistral-7B-v0.1	7.3B	INT4 Q8	54 tok/s 31 tok/s	~3.6s ~3.9s
-	LensVLM 9B	9B	INT4	46 tok/s	~4.4s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4	46 tok/s	~4.4s
-	ChatGLM3-6B-32K	6B	INT4 Q8	44 tok/s 30 tok/s	~3.0s ~3.2s
-	Yi-6B	6B	INT4 Q8	44 tok/s 30 tok/s	~3.0s ~3.2s
-	Falcon3-10B	10B	INT4	42 tok/s	~4.9s
-	Falcon2-11B	11B	INT4	40 tok/s	~5.3s
-	Hunyuan Lite	7B	INT4 Q8	40 tok/s 25 tok/s	~3.4s ~4.1s
-	OLMo 3 7B Base	7B	INT4 Q8	40 tok/s 25 tok/s	~3.4s ~4.1s
-	SEA-LION-7B	7.1B	INT4 Q8	39 tok/s 25 tok/s	~3.5s ~4.1s
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8	39 tok/s 25 tok/s	~3.5s ~4.1s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4	36 tok/s	~3.9s
-	Typhoon-2-8B	8B	INT4	36 tok/s	~3.9s
-	GLM-4-9B	9B	INT4	31 tok/s	~4.8s
-	GLM-4-9B-Chat	9B	INT4	31 tok/s	~4.8s
-	GLM-4-9B-Chat-1M	9B	INT4	31 tok/s	~4.8s
-	Yi-9B	9B	INT4	31 tok/s	~4.8s
-	Sahabat-AI-Gemma2-9B	9.2B	INT4	30 tok/s	~4.9s
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4	30 tok/s	~4.9s

AMD RX 6700 (10GB)

VRAM Scaling by Context Length

Best Models to Run on AMD RX 6700 (10GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions