NVIDIA · consumer

GeForce RTX 3080 Ti Laptop

GeForce RTX 3080 Ti Laptop has 16 GB of VRAM at 512 GB/s — about 14.88 GiB usable after driver and compositor overhead. 1850 of 2118 indexed models fit at 8K context with f16 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
16 GB
GDDR6
Bandwidth
512 GB/s
256-bit bus
Tensor FP16
—
dense
TDP
150 W
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1585vision language 162video 15audio asr 39embedding 26image 2audio tts 21

What fits at 8K context

largest quantization that fits, per model · 1850 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
IQuest-Coder-V1-40B-InstructI1-IQ2_S39.8B11.48 GiB2.50 GiB14.88 GiB0.00 GiB26±12.9%
OLMo-2-1124-13B-InstructQ4_K_M13.7B7.78 GiB6.25 GiB14.88 GiB0.00 GiB26±12.9%
CallerIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
OREAL-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
QwQ-32B-Preview-abliterated-linear25I1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
openhands-lm-32b-v0.1I1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-Coder-32B-abliteratedI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
m1-32bI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
XMainframe-v2-Instruct-32bI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-Coder-32B-Python-SpecialistI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-32b-RP-InkI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
LongWriter-Zero-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
OpenCodeReasoning-Nemotron-32B-IOIIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
OlympicCoder-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
OpenCodeReasoning-Nemotron-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
OpenThinker-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
QwQ-32B-ArliAI-RpR-v4IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-Coder-32B-InstructIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-Coder-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
QwQ-32B-abliteratedIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
DeepSeek-R1-Distill-Qwen-32B-hereticI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
InnoSpark-HPC-RM-32BI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
OpenThinker2-32BIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
INTELLECT-2IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-Coder-32B-Instruct-UncensoredI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
TinyR1-32B-PreviewIQ3_XXS32.8B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Qwen2.5-VL-32B-InstructIQ3_XXS33.5B11.96 GiB2.00 GiB14.86 GiB0.02 GiB26±12.9%
Goetia-26B-A4B-v1.4MoEIQ4_XS26.0B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
G4-Moonlight-Dusk-26B-A4B-hereticMoEIQ4_XS26.5B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
G4-Moonlight-Dusk-26B-A4BMoEIQ4_XS26.5B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
Pantheon-Reasoning-26B-A4B-1.1-hereticMoEIQ4_XS26.5B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
Chimera-X-26B-A4BMoEIQ4_XS26.5B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
Gemma-4-26B-A4B-StyleTune-V2MoEIQ4_XS26.5B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
Gemma-4-26B-A4B-StyleTuneMoEIQ4_XS26.5B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
gemma-4-26b-a4b-heretic-styletune-v2-headMoEIQ4_XS25.8B13.46 GiB0.61 GiB14.86 GiB0.02 GiB26±12.9%
Marco-Mini-InstructMoEI1-Q6_K17.3B13.21 GiB0.88 GiB14.86 GiB0.02 GiB97±37%
cogito-v1-preview-qwen-32BI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.85 GiB0.03 GiB26±12.9%
QwQ-32B-Snowdrop-v0I1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.85 GiB0.03 GiB26±12.9%
DeepSeek-R1-Distill-Qwen-32B-UncensoredI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.85 GiB0.03 GiB26±12.9%
RoguePlanet-DeepSeek-R1-Qwen-32B-RPI1-IQ3_XXS32.8B11.96 GiB2.00 GiB14.85 GiB0.03 GiB26±12.9%
Muse-Glimmer-30BQ3_K_L29.8B13.77 GiB0.20 GiB14.85 GiB0.03 GiB26±12.9%
GPT-NeoX-20B-ErebusI1-IQ2_XS20.6B5.70 GiB8.25 GiB14.84 GiB0.04 GiB26±12.9%
GLM-4.7-Flash-REAP-23B-A3BMoEQ4_123.0B13.62 GiB0.41 GiB14.84 GiB0.04 GiB87±37%
gemma-4-26B-A4B-it-qat-q4_0-unquantizedMoEQ4_026.5B13.45 GiB0.61 GiB14.84 GiB0.04 GiB26±12.9%
diffusiongemma-26B-A4B-itMoENVFP425.8B13.45 GiB0.61 GiB14.84 GiB0.04 GiB26±12.9%
gemma-4-26B-A4BMoEQ4_026.5B13.45 GiB0.61 GiB14.84 GiB0.04 GiB26±12.9%
gemma-4-26B-A4B-Heretic-StableMoEQ4_025.8B13.45 GiB0.61 GiB14.84 GiB0.04 GiB26±12.9%
Magistry-24B-v1.1IQ4_NL23.6B12.67 GiB1.25 GiB14.84 GiB0.04 GiB26±12.9%
Le-Chaton-Slim-23BMoEI1-Q4_K_M23.3B13.21 GiB0.81 GiB14.83 GiB0.05 GiB50±37%
GLM-4.7-FlashMoEQ3_K_M31.2B13.61 GiB0.41 GiB14.83 GiB0.05 GiB98±37%
OpenBuddy-R1-0528-Distill-Qwen3-32B-Preview0-QATIQ3_XXS32.8B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
Qwen3-VL-32B-Instruct-ultra-uncensored-hereticI1-IQ3_XXS33.4B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
Huihui-Qwen3-VL-32B-Instruct-abliteratedI1-IQ3_XXS33.4B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
KAT-DevIQ3_XXS32.8B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
ColorGUI-32BI1-IQ3_XXS33.4B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
Qwen3-VL-32B-InstructIQ3_XXS33.4B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
Qwen3-32B-UncensoredI1-IQ3_XXS32.8B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
Qwen3-32B-abliteratedI1-IQ3_XXS32.8B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
DeepSWE-PreviewIQ3_XXS32.8B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±12.9%
AReaL-boba-2-32BI1-IQ3_XXS32.8B11.94 GiB2.00 GiB14.83 GiB0.05 GiB26±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 generation9.14 it/s6.42–11.63106
Benchmarked· n=106

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 3080 Ti Laptop run?
1850 of 2118 indexed open-weight models fit a GeForce RTX 3080 Ti Laptop at 8,192 context with f16 KV cache, the largest being IQuest-Coder-V1-40B-Instruct at I1-IQ2_S. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 3080 Ti Laptop actually have?
Its nameplate is 16 GB, but about 14.88 GiB is available to a model once driver and compositor overhead is accounted for.
Is a GeForce RTX 3080 Ti Laptop fast for local AI?
Its memory bandwidth is 512 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.