NVIDIA · consumer

GeForce RTX 2080 Ti

GeForce RTX 2080 Ti has 11 GB of VRAM at 616 GB/s — about 10.23 GiB usable after driver and compositor overhead. 839 of 2118 indexed models fit at 64K context with f16 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
11 GB
GDDR6
Bandwidth
616 GB/s
352-bit bus
Tensor FP16
108 TF
dense
TDP
250 W
$999 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 653audio asr 38vision language 92audio tts 20video 14image 1embedding 21

What fits at 64K context

largest quantization that fits, per model · 839 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Aura-4BI1-IQ2_XS4.5B1.41 GiB8.00 GiB10.23 GiB0.00 GiB46±12.9%
magnum-v2-4bI1-IQ2_XS4.5B1.41 GiB8.00 GiB10.23 GiB0.00 GiB46±12.9%
Voxtral-Mini-3B-2507Q3_K_M4.7B1.92 GiB7.50 GiB10.23 GiB0.00 GiB46±12.9%
SuperGemma-4-12b-abliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-uncensored-hereticI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-coder-fable5-composer2.5-v1-uncensored-hereticI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-uncensored-hereticI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Grug-12BI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Aura-Medium-v1-BF16I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-Esper4I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-GuardpointI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Gemma-4-12B-it-AEON-Abliterated-K4-BF16I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-Tachibana-AgentI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-marvin-gutenberg-rp-v2I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-crownelius-writerI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-asterion-agenticI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Huihui-gemma-4-12B-agentic-fable5-abliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
g4-12b-it-trismegistusI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma4-12b-it-asimovI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
FabGemmaI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Huihui-gemma-4-12B-it-qat-q4_0-unquantized-abliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-abliterated-uncensoredI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Gemma-4-12b-it-AbliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-Queen-it-qat-q4_0-unquantizedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-heretic_decensoredI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Iris-12B-gemma-4-it-qatI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-coder-fable5-composer2.5-v1I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
G4-Starry-Ocean-12BI1-IQ3_XS11.9B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-QAT-SOMPOA-heresyI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-uncensored-opus4.7-cotI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Gemma4-12B-IT-AbliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-it-uncensoredI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Huihui-gemma-4-12B-it-abliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-it-hereticI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Tema_Q-X5-12B-ThinkingI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12B-coder-fable5-composer2.5-v1-bf16I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
swarm-sovereign-12bI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Gemma-4-12B-OBLITERATEDI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Gemma4-12B-UncensoredI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Serenity-12BI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Dark-PaneI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Reelva-12BI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
G4-Starry-Ocean-12B-hereticI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Iris-12B-v1.3.2I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Semancer-12BI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Iris-12B-v1.2I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-heretic-abliteratedI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12BIQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-marvin-gutenbergI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-marvin-v2I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
gemma-4-12b-marvin-v1I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
STARK-WEB-12B-v1.7I1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
STARK-WEB-12BI1-IQ3_XS12.0B4.91 GiB4.47 GiB10.22 GiB0.01 GiB47±12.9%
Llama-3.2-3B-Instruct-abliteratedI1-Q5_K_M3.6B2.41 GiB7.00 GiB10.22 GiB0.01 GiB46±12.9%
Llama-3.2-3B-Instruct-uncensoredQ5_K_M3.6B2.41 GiB7.00 GiB10.22 GiB0.01 GiB46±12.9%
Teuken-7B-instruct-research-v0.4Q8_07.5B7.38 GiB2.00 GiB10.22 GiB0.01 GiB47±12.9%
Marco-Nano-InstructMoEI1-IQ1_S8.0B2.44 GiB7.00 GiB10.22 GiB0.01 GiB35±37%
Phi-4-mini-reasoningIQ2_M3.8B1.40 GiB8.00 GiB10.21 GiB0.02 GiB46±12.9%
Phi-4-mini-instructIQ2_M3.8B1.40 GiB8.00 GiB10.21 GiB0.02 GiB46±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.

Measured on this card

third-party benchmarks, aggregated
Workload◍MedianMiddle 50%Runs
Image generation11.62 it/s8.96–13.801,506
Benchmarked· n=1,506

Aggregated from community-submitted runs, so the spread is wide by nature — it covers different models, resolutions, step counts and settings, not one controlled configuration. Read the middle 50% rather than the median alone. These figures are reproduced with attribution from vladmandic-sd-data-benchmark, which publishes no licence — so we display and link rather than redistribute them.

Questions people ask

What AI models can a GeForce RTX 2080 Ti run?
839 of 2118 indexed open-weight models fit a GeForce RTX 2080 Ti at 65,536 context with f16 KV cache, the largest being Aura-4B at I1-IQ2_XS. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 2080 Ti actually have?
Its nameplate is 11 GB, but about 10.23 GiB is available to a model once driver and compositor overhead is accounted for.
Is a GeForce RTX 2080 Ti fast for local AI?
Its memory bandwidth is 616 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.