AMD · consumer

Radeon RX 9060

Radeon RX 9060 has 8 GB of VRAM at 288 GB/s — about 7.44 GiB usable after driver and compositor overhead. 1404 of 2118 indexed models fit at 16K context with q4_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
8 GB
GDDR6
Bandwidth
288 GB/s
128-bit bus
Tensor FP16
—
dense
TDP
132 W
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1208vision language 102embedding 26video 7audio tts 21image 2audio asr 38

What fits at 16K context

largest quantization that fits, per model · 1404 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Smilodon-9B-v1I1-Q4_110.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
bella-bartender-v2I1-Q4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
Gemma-The-Writer-9B-HERETIC-Uncensored-AbliteratedI1-Q4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFWI1-Q4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
Gemma-SEA-LION-v3-9B-ITI1-Q4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
G2-Darkest-Writer-Dirty-Shirley-9B-v2I1-Q4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
G2-Darkest-Writer-9B-v1I1-Q4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
gemma-2-9bQ4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
gemma-2-9b-itQ4_19.2B5.55 GiB0.95 GiB7.44 GiB0.00 GiB28±26.5%
HomunculusQ3_K_M12.5B5.79 GiB0.70 GiB7.44 GiB0.00 GiB28±26.5%
SOLAR-10.7B-Instruct-v1.0-uncensoredQ4_010.7B5.66 GiB0.84 GiB7.44 GiB0.00 GiB28±26.5%
Nous-Hermes-2-SOLAR-10.7BQ4_010.7B5.66 GiB0.84 GiB7.44 GiB0.00 GiB28±26.5%
SOLAR-10.7B-Instruct-v1.0Q4_010.7B5.66 GiB0.84 GiB7.44 GiB0.00 GiB28±26.5%
salamandra-7b-instruct-2606I1-Q6_K7.8B5.94 GiB0.56 GiB7.43 GiB0.01 GiB28±26.5%
v6-Finch-7B-HFIQ4_XS7.6B4.24 GiB2.25 GiB7.43 GiB0.01 GiB28±26.5%
rwkv-6-world-7bIQ4_XS7.6B4.24 GiB2.25 GiB7.43 GiB0.01 GiB28±26.5%
LFM2.5-Queen-Opus-4.7-8B-A1BMoEI1-Q6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
LFM2.5-8B-A1B-KO-SFTMoEI1-Q6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
LFM2.5-8B-A1B-hereticMoEI1-Q6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
LFM2.5-8B-A1B-SOMPOA-heresyMoEI1-Q6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
Huihui-LFM2.5-8B-A1B-abliteratedMoEI1-Q6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
Supertron2.1-8B-A1BMoEI1-Q6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
LFM2.5-8B-A1BMoEQ6_K8.5B6.48 GiB0.05 GiB7.43 GiB0.01 GiB78±37%
Gemma-4-12B-StyleTuneI1-Q3_K_M13.0B6.07 GiB0.41 GiB7.43 GiB0.01 GiB28±26.5%
gemma-4-12b-heretic-styletune-headI1-Q3_K_M12.0B6.07 GiB0.41 GiB7.43 GiB0.01 GiB28±26.5%
syrian-gemma-12bI1-Q3_K_M13.0B6.07 GiB0.41 GiB7.43 GiB0.01 GiB28±26.5%
Qwen3.6-35B-A3B-REAM-160-ru-agentMoEIQ2_XXS23.6B6.43 GiB0.09 GiB7.43 GiB0.01 GiB114±37%
DeepSeek-R1-Distill-Llama-8B-AbliteratedI1-Q2_K8.0B5.92 GiB0.56 GiB7.42 GiB0.02 GiB28±26.5%
codegeex4-all-9bIQ2_M9.4B3.66 GiB2.81 GiB7.42 GiB0.02 GiB28±26.5%
Ling-mini-2.0MoEIQ3_XS16.3B6.35 GiB0.18 GiB7.42 GiB0.02 GiB118±37%
glm-4-9b-chat-abliteratedIQ2_M9.4B3.66 GiB2.81 GiB7.42 GiB0.02 GiB28±26.5%
glm-4-9b-chatIQ2_M9.4B3.66 GiB2.81 GiB7.42 GiB0.02 GiB28±26.5%
Wan2.1-T2V-1.3BQ4_01.4B6.50 GiB0.00 GiB7.42 GiB0.02 GiB28±26.5%
Qwythos-9B-v2Q4_K_L9.7B6.34 GiB0.14 GiB7.41 GiB0.03 GiB28±26.5%
Tess-4-9BQ4_K_L9.7B6.34 GiB0.14 GiB7.41 GiB0.03 GiB28±26.5%
Qwen3.5-9B-BaseQ5_19.7B6.33 GiB0.14 GiB7.41 GiB0.03 GiB28±26.5%
Teuken-7B-instruct-research-v0.4Q6_K_L7.5B6.33 GiB0.14 GiB7.41 GiB0.03 GiB28±26.5%
LocateAnything-3BBF163.8B6.34 GiB0.16 GiB7.41 GiB0.03 GiB28±26.5%
Snowpiercer-15B-v4-hereticI1-IQ3_XXS15.0B5.58 GiB0.88 GiB7.41 GiB0.03 GiB28±26.5%
Snowpiercer-15B-v4IQ3_XXS15.0B5.58 GiB0.88 GiB7.41 GiB0.03 GiB28±26.5%
Qwen2.5-3B-Instruct-abliteratedF163.1B6.33 GiB0.16 GiB7.40 GiB0.04 GiB28±26.5%
GRM-Kerlin-3bF163.4B6.33 GiB0.16 GiB7.40 GiB0.04 GiB28±26.5%
Garnet-OCR-3B-0422F164.1B6.33 GiB0.16 GiB7.40 GiB0.04 GiB28±26.5%
Qwen3-16B-A3BMoEIQ3_XXS16.0B6.08 GiB0.42 GiB7.40 GiB0.04 GiB64±37%
gemma-3-12b-it-vl-Gemini-3-Pro-Preview-Heretic-Uncensored-ThinkingI1-Q3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-it-vl-Deepseek-v3.1-Heretic-Uncensored-ThinkingI1-Q3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-it-ultra-uncensored-hereticQ3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-it-vl-GLM-4.7-Flash-Heretic-Uncensored-ThinkingI1-Q3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
Floppa-12B-Gemma3-UncensoredI1-Q3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-it-hereticI1-Q3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-it-abliteratedQ3_K_L12.2B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-it-abliterated-v2Q3_K_L11.8B6.04 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
gemma-3-12b-itQ3_K_L12.2B6.03 GiB0.41 GiB7.39 GiB0.05 GiB28±26.5%
Dolphin3.0-Llama3.2-3BF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
Llama-Song-Stream-3B-InstructF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
Llama-3.2-3B-Instruct-roleplay-tunedF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
Llama-Doctor-3.2-3B-InstructF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
Llama-3.2-3B-Instruct-uncensoredF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
Llama-3.2-3B-Instruct-heretic-ablitered-uncensoredF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
llama-3.2-Korean-Bllossom-3BF163.2B5.99 GiB0.49 GiB7.39 GiB0.05 GiB28±26.5%
●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a Radeon RX 9060 run?
1404 of 2118 indexed open-weight models fit a Radeon RX 9060 at 16,384 context with q4_0 KV cache, the largest being Smilodon-9B-v1 at I1-Q4_1. That covers text, vision-language, image, video and speech models.
How much usable memory does a Radeon RX 9060 actually have?
Its nameplate is 8 GB, but about 7.44 GiB is available to a model once driver and compositor overhead is accounted for.
Is a Radeon RX 9060 fast for local AI?
Its memory bandwidth is 288 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.