Skip to content

AMD Radeon RX 6700 XT

AMD · 12GBGDDR6 · Can run 63 models

BuyAmazon
ManufacturerAMD
VRAM12 GB
Memory TypeGDDR6
ArchitectureRDNA 2
Stream Procs2,560
Bandwidth384 GB/s
TDP230W
MSRP$479
ReleasedMar 18, 2021

AI Notes

The RX 6700 XT offers 12GB VRAM, enough to run 7B models at higher quantization levels and 13B models with Q4 quantization. ROCm support for RDNA 2 is available but requires manual setup. A solid budget option on the used market for those willing to work with the AMD software stack.

Compatible Models

ModelParametersBest QuantVRAM UsedFitEst. Speed
Qwen 3 0.6B600MQ4_K_M2.5 GBRuns~154 tok/s
Qwen 3.5 0.8B800MQ4_K_M1.5 GBRuns~256 tok/s
Gemma 3 1B1BQ8_02 GBRuns~192 tok/s
Llama 3.2 1B1BQ8_03 GBRuns~128 tok/s
DeepSeek R1 1.5B1.5BQ8_03 GBRuns~128 tok/s
SmolLM2 1.7B1.7BQ8_02.7 GBRuns~142 tok/s
Gemma 2 2B2BQ8_04 GBRuns~96 tok/s
Gemma 3n E2B2BQ4_K_M3.3 GBRuns~116 tok/s
Gemma 4 E2B2BQ4_K_M4 GBRuns~96 tok/s
Qwen 3.5 2B2BQ4_K_M3 GBRuns~128 tok/s
Llama 3.2 3B3BQ8_05 GBRuns~77 tok/s
StarCoder2 3B3BQ4_K_M3.5 GBRuns~110 tok/s
Phi-3 Mini 3.8B3.8BQ8_05.8 GBRuns~66 tok/s
Phi-4 Mini 3.8B3.8BQ4_K_M4.5 GBRuns~85 tok/s
Gemma 3 4B4BQ4_K_M5 GBRuns~77 tok/s
Gemma 3n E4B4BQ4_K_M4.5 GBRuns~85 tok/s
Gemma 4 E4B4BQ4_K_M6 GBRuns~64 tok/s
Qwen 3 4B4BQ4_K_M4.5 GBRuns~85 tok/s
Qwen 3.5 4B4BQ4_K_M4.5 GBRuns~85 tok/s
Yi 1.5 6B6BQ4_K_M5 GBRuns~77 tok/s
Codestral Mamba 7B7BQ4_K_M6.9 GBRuns~56 tok/s
DeepSeek R1 7B7BQ8_09 GBRuns~43 tok/s
Falcon 3 7B7BQ4_K_M6.8 GBRuns~56 tok/s
InternLM 2.5 7B7BQ4_K_M5.5 GBRuns~70 tok/s
Mistral 7B7BQ8_09 GBRuns~43 tok/s
OpenChat 3.5 7B7BQ4_K_M6.9 GBRuns~56 tok/s
Qwen 2.5 7B7BQ8_09 GBRuns~43 tok/s
Qwen 2.5 Coder 7B7BQ8_09 GBRuns~43 tok/s
Qwen 2.5 VL 7B7BQ4_K_M7 GBRuns~55 tok/s
StarCoder2 7B7BQ4_K_M5.5 GBRuns~70 tok/s
WizardLM 2 7B7BQ4_K_M6.9 GBRuns~56 tok/s
Aya Expanse 8B8BQ4_K_M6.5 GBRuns~59 tok/s
Cogito 8B8BQ4_K_M7.5 GBRuns~51 tok/s
DeepSeek R1 8B8BQ4_K_M7.5 GBRuns~51 tok/s
Dolphin 3 8B8BQ4_K_M6 GBRuns~64 tok/s
Granite 3.3 8B8BQ8_010 GBRuns~38 tok/s
Llama 3.1 8B8BQ8_010 GBRuns~38 tok/s
Nemotron 3 Nano 8B8BQ4_K_M7.5 GBRuns~51 tok/s
Nous Hermes 2 8B8BQ4_K_M6 GBRuns~64 tok/s
Qwen 3 8B8BQ4_K_M7.5 GBRuns~51 tok/s
Qwen 3.5 9B9BQ4_K_M7.5 GBRuns~51 tok/s
Yi 1.5 9B9BQ4_K_M6.5 GBRuns~59 tok/s
Yi Coder 9B9BQ4_K_M8 GBRuns~48 tok/s
Falcon 3 10B10BQ4_K_M8.5 GBRuns~45 tok/s
Llama 3.2 Vision 11B11BQ4_K_M8.5 GBRuns~45 tok/s
Mistral Nemo 12B12BQ4_K_M9.5 GBRuns~40 tok/s
DeepSeek R1 14B14BQ4_K_M9.9 GBRuns~39 tok/s
Phi-4 14B14BQ4_K_M9.9 GBRuns~39 tok/s
Qwen 2.5 14B14BQ4_K_M9.9 GBRuns~39 tok/s
Gemma 2 9B9BQ8_011 GBRuns (tight)~35 tok/s
Gemma 3 12B12BQ4_K_M10.5 GBRuns (tight)~37 tok/s
Phi-4 Reasoning 14B14BQ4_K_M11 GBRuns (tight)~35 tok/s
Qwen 2.5 Coder 14B14BQ4_K_M12 GBCPU Offload~10 tok/s
Qwen 3 14B14BQ4_K_M12 GBCPU Offload~10 tok/s
StarCoder2 15B15BQ8_017 GBCPU Offload~7 tok/s
InternLM 2.5 20B20BQ4_K_M12 GBCPU Offload~10 tok/s
gpt-oss 20B21BMXFP415 GBCPU Offload~8 tok/s
Codestral 22B22BQ4_K_M14.7 GBCPU Offload~8 tok/s
Devstral 24B24BQ4_K_M17 GBCPU Offload~7 tok/s
Magistral Small 24B24BQ4_K_M17 GBCPU Offload~7 tok/s
Mistral Small 3.1 24B24BQ4_K_M18 GBCPU Offload~6 tok/s
Gemma 2 27B27BQ4_K_M17.7 GBCPU Offload~7 tok/s
Qwen 3.5 35B A3B35BQ4_K_M12 GBCPU Offload~10 tok/s
51 model(s) are too large for this hardware.