Skip to content

AMD Radeon RX 6800 XT

AMD · 16GBGDDR6 · Can run 81 models

BuyAmazon
ManufacturerAMD
VRAM16 GB
Memory TypeGDDR6
ArchitectureRDNA 2
Stream Procs4,608
Bandwidth512 GB/s
TDP300W
MSRP$649
ReleasedNov 18, 2020

AI Notes

The RX 6800 XT offers 16GB VRAM and 512 GB/s bandwidth, making it a strong option for local AI on the used market. It runs 13B models comfortably and can attempt 30B with quantization. ROCm support for RDNA 2 is mature, though setup still requires more effort than NVIDIA's CUDA ecosystem.

Compatible Models

ModelParametersBest QuantVRAM UsedFitEst. Speed
Qwen 3 0.6B600MQ4_K_M2.5 GBRuns~205 tok/s
Qwen 3.5 0.8B800MQ4_K_M1.5 GBRuns~341 tok/s
Gemma 3 1B1BQ8_02 GBRuns~256 tok/s
Llama 3.2 1B1BQ8_03 GBRuns~171 tok/s
DeepSeek R1 1.5B1.5BQ8_03 GBRuns~171 tok/s
SmolLM2 1.7B1.7BQ8_02.7 GBRuns~190 tok/s
Gemma 2 2B2BQ8_04 GBRuns~128 tok/s
Gemma 3n E2B2BQ4_K_M3.3 GBRuns~155 tok/s
Gemma 4 E2B2BQ4_K_M4 GBRuns~128 tok/s
Qwen 3.5 2B2BQ4_K_M3 GBRuns~171 tok/s
Llama 3.2 3B3BQ8_05 GBRuns~102 tok/s
StarCoder2 3B3BQ4_K_M3.5 GBRuns~146 tok/s
Phi-3 Mini 3.8B3.8BQ8_05.8 GBRuns~88 tok/s
Phi-4 Mini 3.8B3.8BQ4_K_M4.5 GBRuns~114 tok/s
Gemma 3 4B4BQ4_K_M5 GBRuns~102 tok/s
Gemma 3n E4B4BQ4_K_M4.5 GBRuns~114 tok/s
Gemma 4 E4B4BQ4_K_M6 GBRuns~85 tok/s
Qwen 3 4B4BQ4_K_M4.5 GBRuns~114 tok/s
Qwen 3.5 4B4BQ4_K_M4.5 GBRuns~114 tok/s
Yi 1.5 6B6BQ4_K_M5 GBRuns~102 tok/s
Codestral Mamba 7B7BQ4_K_M6.9 GBRuns~74 tok/s
DeepSeek R1 7B7BQ8_09 GBRuns~57 tok/s
Falcon 3 7B7BQ4_K_M6.8 GBRuns~75 tok/s
InternLM 2.5 7B7BQ4_K_M5.5 GBRuns~93 tok/s
Mistral 7B7BQ8_09 GBRuns~57 tok/s
OpenChat 3.5 7B7BQ4_K_M6.9 GBRuns~74 tok/s
Qwen 2.5 7B7BQ8_09 GBRuns~57 tok/s
Qwen 2.5 Coder 7B7BQ8_09 GBRuns~57 tok/s
Qwen 2.5 VL 7B7BQ4_K_M7 GBRuns~73 tok/s
StarCoder2 7B7BQ4_K_M5.5 GBRuns~93 tok/s
WizardLM 2 7B7BQ4_K_M6.9 GBRuns~74 tok/s
Aya Expanse 8B8BQ4_K_M6.5 GBRuns~79 tok/s
Cogito 8B8BQ4_K_M7.5 GBRuns~68 tok/s
DeepSeek R1 8B8BQ4_K_M7.5 GBRuns~68 tok/s
Dolphin 3 8B8BQ4_K_M6 GBRuns~85 tok/s
Granite 3.3 8B8BQ8_010 GBRuns~51 tok/s
Llama 3.1 8B8BQ8_010 GBRuns~51 tok/s
Nemotron 3 Nano 8B8BQ4_K_M7.5 GBRuns~68 tok/s
Nous Hermes 2 8B8BQ4_K_M6 GBRuns~85 tok/s
Qwen 3 8B8BQ4_K_M7.5 GBRuns~68 tok/s
Gemma 2 9B9BQ8_011 GBRuns~47 tok/s
Qwen 3.5 9B9BQ4_K_M7.5 GBRuns~68 tok/s
Yi 1.5 9B9BQ4_K_M6.5 GBRuns~79 tok/s
Yi Coder 9B9BQ4_K_M8 GBRuns~64 tok/s
Falcon 3 10B10BQ4_K_M8.5 GBRuns~60 tok/s
Llama 3.2 Vision 11B11BQ4_K_M8.5 GBRuns~60 tok/s
Gemma 3 12B12BQ4_K_M10.5 GBRuns~49 tok/s
Mistral Nemo 12B12BQ4_K_M9.5 GBRuns~54 tok/s
DeepSeek R1 14B14BQ4_K_M9.9 GBRuns~52 tok/s
Phi-4 14B14BQ4_K_M9.9 GBRuns~52 tok/s
Phi-4 Reasoning 14B14BQ4_K_M11 GBRuns~47 tok/s
Qwen 2.5 14B14BQ4_K_M9.9 GBRuns~52 tok/s
Qwen 2.5 Coder 14B14BQ4_K_M12 GBRuns~43 tok/s
Qwen 3 14B14BQ4_K_M12 GBRuns~43 tok/s
InternLM 2.5 20B20BQ4_K_M12 GBRuns~43 tok/s
Qwen 3.5 35B A3B35BQ4_K_M12 GBRuns~43 tok/s
gpt-oss 20B21BMXFP415 GBRuns (tight)~34 tok/s
Codestral 22B22BQ4_K_M14.7 GBRuns (tight)~35 tok/s
StarCoder2 15B15BQ8_017 GBCPU Offload~9 tok/s
Devstral 24B24BQ4_K_M17 GBCPU Offload~9 tok/s
Magistral Small 24B24BQ4_K_M17 GBCPU Offload~9 tok/s
Mistral Small 3.1 24B24BQ4_K_M18 GBCPU Offload~8 tok/s
Gemma 4 26B26BQ4_K_M20 GBCPU Offload~8 tok/s
Gemma 2 27B27BQ4_K_M17.7 GBCPU Offload~9 tok/s
Gemma 3 27B27BQ4_K_M20 GBCPU Offload~8 tok/s
Qwen 3.5 27B27BQ4_K_M19 GBCPU Offload~8 tok/s
Qwen 3.6 27B27BQ4_K_M20 GBCPU Offload~8 tok/s
Qwen 3 30B-A3B (MoE)30BQ4_K_M22 GBCPU Offload~7 tok/s
Gemma 4 31B31BQ4_K_M22 GBCPU Offload~7 tok/s
Aya Expanse 32B32BQ4_K_M22 GBCPU Offload~7 tok/s
Cogito 32B32BQ4_K_M21.5 GBCPU Offload~7 tok/s
DeepSeek R1 32B32BQ4_K_M20.7 GBCPU Offload~8 tok/s
Qwen 2.5 32B32BQ4_K_M20.7 GBCPU Offload~8 tok/s
Qwen 2.5 Coder 32B32BQ4_K_M23 GBCPU Offload~7 tok/s
Qwen 3 32B32BQ4_K_M23 GBCPU Offload~7 tok/s
QwQ 32B32BQ4_K_M21.5 GBCPU Offload~7 tok/s
Laguna XS 2.133BQ4_K_M22 GBCPU Offload~7 tok/s
WizardCoder 33B33BQ4_K_M22 GBCPU Offload~7 tok/s
Nous Hermes 2 34B34BQ4_K_M19 GBCPU Offload~8 tok/s
Yi 1.5 34B34BQ4_K_M21 GBCPU Offload~7 tok/s
Command R 35B35BQ4_K_M22.5 GBCPU Offload~7 tok/s
33 model(s) are too large for this hardware.