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. 1793 of 2118 indexed models fit at 128K context with q4_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
vision language 170text 1520audio asr 39video 16image 1embedding 26audio tts 21

What fits at 128K context

largest quantization that fits, per model · 1793 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Gemma4-Gutenberg-31BQ2_K31.3B11.76 GiB5.95 GiB18.60 GiB0.00 GiB31±12.9%
gemma-4-31B-itQ2_K31.3B11.76 GiB5.95 GiB18.60 GiB0.00 GiB31±12.9%
Gemma4-Gutenberg-31B-HereticQ2_K31.3B11.76 GiB5.95 GiB18.60 GiB0.00 GiB31±12.9%
Equinox-31BQ2_K31.3B11.76 GiB5.95 GiB18.60 GiB0.00 GiB31±12.9%
gemma-4-31B-it-SDFT-Heretic-RPQ2_K30.7B11.76 GiB5.95 GiB18.60 GiB0.00 GiB31±12.9%
QwQ-32BUD-IQ2_XXS32.8B8.70 GiB9.00 GiB18.59 GiB0.01 GiB31±12.9%
North-Mini-Code-1.0MoEQ4_K_S30.5B16.81 GiB1.00 GiB18.59 GiB0.01 GiB98±37%
t5-v1_1-xxlF324.8B17.74 GiB0.00 GiB18.59 GiB0.01 GiB31±12.9%
Qwen3-42B-A3B-2507-Thinking-Abliterated-uncensored-TOTAL-RECALL-v2-Medium-MASTER-CODERMoEI1-IQ2_M42.4B13.08 GiB4.71 GiB18.59 GiB0.01 GiB49±37%
Qwen3-Coder-Next-Opus-4.6-Reasoning-DistilledMoEIQ1_M—16.95 GiB0.84 GiB18.59 GiB0.01 GiB132±37%
gemma-4-26B-A4BMoEQ5_026.5B16.30 GiB1.49 GiB18.57 GiB0.03 GiB31±12.9%
gemma-4-26B-A4B-Heretic-StableMoEQ5_025.8B16.30 GiB1.49 GiB18.57 GiB0.03 GiB31±12.9%
GLM-4.7-Flash-DerestrictedMoEI1-Q4_K_S31.2B15.90 GiB1.86 GiB18.57 GiB0.03 GiB80±37%
Huihui-GLM-4.7-Flash-abliteratedMoEI1-Q4_K_S31.2B15.90 GiB1.86 GiB18.57 GiB0.03 GiB80±37%
Qwen3.5-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-ThinkingI1-IQ3_XXS39.5B14.33 GiB3.38 GiB18.57 GiB0.03 GiB31±12.9%
Magistry-24B-v1.1IQ4_XS23.6B12.02 GiB5.63 GiB18.56 GiB0.04 GiB31±12.9%
dolphin-2.9.2-Phi-3-MediumKV unresolvedQ6_K14.0B10.67 GiB7.03 GiB18.56 GiB0.04 GiB31±12.9%
Phi-3-medium-128k-instructQ6_K14.0B10.67 GiB7.03 GiB18.56 GiB0.04 GiB31±12.9%
Phi-3-medium-4k-instructI1-Q6_K14.0B10.67 GiB7.03 GiB18.56 GiB0.04 GiB31±12.9%
Llama3.2-24B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-UncensoredMoEI1-Q6_K18.0B13.81 GiB3.94 GiB18.56 GiB0.04 GiB48±37%
L3-DARKEST-PLANET-16.5BQ3_K_M16.5B7.73 GiB9.98 GiB18.55 GiB0.05 GiB31±12.9%
ALIA-40b-fc-2606I1-IQ2_XXS40.4B10.89 GiB6.75 GiB18.55 GiB0.05 GiB31±12.9%
ALIA-40b-instruct-2606I1-IQ2_XXS40.4B10.89 GiB6.75 GiB18.55 GiB0.05 GiB31±12.9%
INTELLECT-2UD-IQ2_XXS32.8B8.65 GiB9.00 GiB18.55 GiB0.05 GiB31±12.9%
Nous-Hermes-2-Yi-34BI1-IQ2_XXS34.4B9.23 GiB8.44 GiB18.55 GiB0.05 GiB31±12.9%
Nous-Capybara-limarpv3-34BI1-IQ2_XXS34.4B9.23 GiB8.44 GiB18.55 GiB0.05 GiB31±12.9%
Kimi-Linear-48B-A3B-InstructMoEQ2_K_L49.1B16.67 GiB1.07 GiB18.55 GiB0.05 GiB31±12.9%
Transformed-Journey-24BIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Mergedonia-AETHER-24B-v1aIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Mergedonia-AETHER-24B-v1bIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Slimaki-Tavern-24B-v1.3IQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Maginum-Cydoms-24BIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Maginum-Cydoms-24B-absolute-heresyIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Mistral-Small-3.2-24B-Instruct-2506-ultra-uncensored-hereticIQ4_XS24.0B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Huihui-Mistral-Small-3.2-24B-Instruct-2506-abliterated-llamacppfixedIQ4_XS24.0B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Mistral-Small-3_2-24B-Instruct-2506-antislop.v2IQ4_XS24.0B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Goetia-24B-v1.1IQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
MagiSeek-Pro-V1IQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Cogidonia-v2-24BIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Precog-24B-v1IQ4_XS—12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
experiment024bIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Cydonia-24B-v4.3-absolute-heresyIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Cydonia-24B-v4.3-heretic-v2IQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Cydonia-24B-v4.3-hereticIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Cydonia-24B-v4.3-heretic-v4IQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Journeys-End-24BIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Dolphin-Mistral-GLM-4.7-Flash-24B-Venice-Edition-Thinking-UncensoredIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
WeirdCompound-v1.7-24bIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
Mistral-Small-24B-Instruct-JbliteratedIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
grok-oss-Apollyon-24BIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
grok-oss-Apollyon-24B-hereticIQ4_XS23.6B12.00 GiB5.63 GiB18.55 GiB0.05 GiB31±12.9%
UncensoredLM-DeepSeek-R1-Distill-Qwen-14BQ6_K_L14.2B11.22 GiB6.47 GiB18.54 GiB0.06 GiB31±12.9%
Pantheon-Reasoning-27BQ4_027.8B15.42 GiB2.25 GiB18.54 GiB0.06 GiB31±12.9%
Qwen3.5-27BQ4_027.8B15.42 GiB2.25 GiB18.54 GiB0.06 GiB31±12.9%
Qwen3-Coder-REAP-25B-A3BMoEQ4_K_L24.9B14.37 GiB3.38 GiB18.53 GiB0.07 GiB58±37%
Qwen3-VL-32B-InstructUD-IQ2_XXS33.4B8.64 GiB9.00 GiB18.53 GiB0.07 GiB31±12.9%
Qwen3-VL-32B-ThinkingUD-IQ2_XXS33.4B8.64 GiB9.00 GiB18.53 GiB0.07 GiB31±12.9%
Qwen3-32BUD-IQ2_XXS32.8B8.64 GiB9.00 GiB18.53 GiB0.07 GiB31±12.9%
Ornith-1.0-35B-uncensored-hereticMoEQ3_K_L35.1B17.02 GiB0.70 GiB18.53 GiB0.07 GiB129±37%
Pantheon-Reasoning-26B-A4B-1.1MoEQ4_K_M26.5B16.25 GiB1.49 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?
1793 of 2118 indexed open-weight models fit a GeForce RTX 3080 Ti at 131,072 context with q4_0 KV cache, the largest being Gemma4-Gutenberg-31B at Q2_K. 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.