NVIDIA · consumer

GeForce RTX 3080 Ti

GeForce RTX 3080 Ti has 20 GB of VRAM at 760 GB/s — about 18.60 GiB usable after driver and compositor overhead. 1946 of 2118 indexed models fit at 8K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
20 GB
GDDR6X
Bandwidth
760 GB/s
320-bit bus
Tensor FP16
136 TF
dense
TDP
350 W
$1199 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1671vision language 171video 16audio asr 39image 2audio tts 21embedding 26

What fits at 8K context

largest quantization that fits, per model · 1946 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Skyfall-31B-v4.2Q4_031.4B16.78 GiB0.90 GiB18.60 GiB0.00 GiB31±12.9%
reka-flash-3.1I1-Q6_K20.9B17.17 GiB0.55 GiB18.60 GiB0.00 GiB31±12.9%
reka-flash-3Q6_K20.9B17.17 GiB0.55 GiB18.60 GiB0.00 GiB31±12.9%
Huihui-gemma-4-26B-A4B-it-abliteratedMoEUD-Q5_K_S26.5B17.48 GiB0.32 GiB18.59 GiB0.01 GiB31±12.9%
t5-v1_1-xxlF324.8B17.74 GiB0.00 GiB18.59 GiB0.01 GiB31±12.9%
OpenBuddy-R1-0528-Distill-Qwen3-32B-Preview0-QATIQ4_XS32.8B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
Qwen3-VL-32B-Instruct-ultra-uncensored-hereticIQ4_XS33.4B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
KAT-DevIQ4_XS32.8B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
Huihui-Qwen3-VL-32B-Instruct-abliteratedIQ4_XS33.4B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
ColorGUI-32BIQ4_XS33.4B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
Qwen3-32B-UncensoredIQ4_XS32.8B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
Qwen3-32B-abliteratedIQ4_XS32.8B16.63 GiB1.06 GiB18.58 GiB0.02 GiB31±12.9%
AMALIA-9B-0626-DPOBF169.2B17.05 GiB0.70 GiB18.58 GiB0.02 GiB31±12.9%
EXAONE-4.0-32BQ4_032.0B16.96 GiB0.71 GiB18.57 GiB0.03 GiB31±12.9%
Gemma-4-31B-Isometry-RPIQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
Gemma-4-Dark-Gemistry-31BIQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
Prosopon-31BIQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
Giftige-Blume-31B-v1-StyleSwapIQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
G4-MeroMero-31B-StyleSwapIQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
Gemma-4-31B-StyleTune-heretic-araIQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
Pantheon-Reasoning-31B-1.1IQ4_XS32.7B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
Barcenas-StyleTune-31B-FableIQ4_XS32.1B16.40 GiB1.29 GiB18.57 GiB0.03 GiB31±12.9%
GLM-Z1-32B-0414-uncensored-heretic-v2Q4_K_S32.6B17.42 GiB0.25 GiB18.57 GiB0.03 GiB31±12.9%
Wizard-Vicuna-30B-UncensoredI1-Q2_K32.5B11.22 GiB6.47 GiB18.57 GiB0.03 GiB31±12.9%
archangel_sft-kto_llama30bI1-Q2_K32.5B11.22 GiB6.47 GiB18.57 GiB0.03 GiB31±12.9%
Qwen3.6-34B-80L-Fable-5-HereticI1-IQ4_XS33.4B17.37 GiB0.33 GiB18.56 GiB0.04 GiB31±12.9%
GRM-2.6-Plus-0628Q4_K_L27.8B17.43 GiB0.27 GiB18.56 GiB0.04 GiB31±12.9%
ThinkingCap-Qwen3.6-27BQ4_K_L27.4B17.43 GiB0.27 GiB18.56 GiB0.04 GiB31±12.9%
Tess-4-27BQ4_K_L27.8B17.43 GiB0.27 GiB18.56 GiB0.04 GiB31±12.9%
GLM-4-32B-0414-Korean-CultureI1-Q4_K_S32.6B17.41 GiB0.25 GiB18.56 GiB0.04 GiB31±12.9%
GLM-Z1-32B-0414Q4_K_S32.6B17.41 GiB0.25 GiB18.56 GiB0.04 GiB31±12.9%
GLM-4-32B-0414Q4_K_S32.6B17.41 GiB0.25 GiB18.56 GiB0.04 GiB31±12.9%
Qwen3-Coder-30B-A3B-InstructMoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
Qwen3-VL-30B-A3B-ThinkingMoEQ4_K_M31.1B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
MiroThinker-v1.0-30BMoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
Qwen3-30B-A3BMoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
Qwen3-30B-A3B-Instruct-2507MoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
Qwen3-30B-A3B-Thinking-2507MoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
Pantheon-Proto-RP-1.8-30B-A3BMoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
GLM-4.7-Flash-DerestrictedMoEI1-Q4_131.2B17.52 GiB0.22 GiB18.54 GiB0.06 GiB126±37%
Huihui-GLM-4.7-Flash-abliteratedMoEI1-Q4_131.2B17.52 GiB0.22 GiB18.54 GiB0.06 GiB126±37%
Tongyi-DeepResearch-30B-A3BMoEQ4_K_M30.5B17.35 GiB0.40 GiB18.54 GiB0.06 GiB118±37%
Huihui-GLM-4.7-Flash-abliterated-57BMoEI1-IQ2_M57.3B17.14 GiB0.56 GiB18.54 GiB0.06 GiB113±37%
Skyfall-31B-v4.2-hereticI1-Q4_K_S31.4B16.71 GiB0.90 GiB18.53 GiB0.07 GiB31±12.9%
Holo3-35B-A3BMoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen35B-Agent-R2-AbliteratedMoEIQ4_XS34.7B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Darwin-35B-A3B-OpusMoEIQ4_XS36.0B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
spoomplesmaxx-flash-35B-A3MoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Holo-3.1-35B-A3BMoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
grug-35b-v2MoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen3.6-35B-A3B-abliterated-MAXMoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen3.6-35B-A3B-StyleTuneMoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen3.6-35B-A3B-Kimi-K2.6-Reasoning-DistilledMoEIQ4_XS36.0B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-DistilledMoEIQ4_XS36.0B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen3.6-35B-A3B-abliteratedMoEIQ4_XS35.1B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliteratedMoEIQ4_XS36.0B17.64 GiB0.08 GiB18.53 GiB0.07 GiB168±37%
Qwen3.6-27B-uncensored-heretic-v2Q5_K_S27.4B17.40 GiB0.27 GiB18.53 GiB0.07 GiB31±12.9%
Qwen3.5-27B-Engineer-Deckard-GeminiI1-Q5_K_S27.7B17.40 GiB0.27 GiB18.53 GiB0.07 GiB31±12.9%
Qwen3.5-27B-HERETIC-Polaris-Advanced-Thinking-Alpha-uncensoredI1-Q5_K_S27.4B17.40 GiB0.27 GiB18.53 GiB0.07 GiB31±12.9%
Qwen3.5-27B-Deckard-PKD-Heretic-Uncensored-ThinkingI1-Q5_K_S27.4B17.40 GiB0.27 GiB18.53 GiB0.07 GiB31±12.9%
●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 GeForce RTX 3080 Ti run?
1946 of 2118 indexed open-weight models fit a GeForce RTX 3080 Ti at 8,192 context with q8_0 KV cache, the largest being Skyfall-31B-v4.2 at Q4_0. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 3080 Ti actually have?
Its nameplate is 20 GB, but about 18.60 GiB is available to a model once driver and compositor overhead is accounted for.
Is a GeForce RTX 3080 Ti fast for local AI?
Its memory bandwidth is 760 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.