NVIDIA · datacenter

Tesla V100 32GB

Tesla V100 32GB has 32 GB of VRAM at 900 GB/s — about 29.76 GiB usable after driver and compositor overhead. 1859 of 2118 indexed models fit at 128K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
32 GB
HBM2
Bandwidth
900 GB/s
4096-bit bus
Tensor FP16
125 TF
dense
TDP
300 W
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
vision language 176text 1580video 16audio tts 21image 1audio asr 39embedding 26

What fits at 128K context

largest quantization that fits, per model · 1859 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Qwen3-VL-32B-ThinkingQ2_K_L33.4B11.67 GiB17.00 GiB29.76 GiB0.00 GiB18±22%
Qwen3-32BQ2_K_L32.8B11.67 GiB17.00 GiB29.76 GiB0.00 GiB18±22%
Bernini-RQ8_014.3B28.71 GiB0.00 GiB29.76 GiB0.00 GiB18±22%
Qwen3.5-88BMoEI1-IQ2_M87.7B27.12 GiB1.59 GiB29.74 GiB0.02 GiB65±37%
QwQ-32BQ2_K_L32.8B11.64 GiB17.00 GiB29.74 GiB0.02 GiB18±22%
NVIDIA-Nemotron-Nano-12B-v2Q8_012.3B12.19 GiB16.47 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Gembrain-X-Core-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Gembrain-X-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-31B-Isometry-Fabled-PersonaI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Versipellis-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma4-Gutenberg-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
G4-MeroMero-31B-uncensored-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Novelist-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Wanabi-Gemma4-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
G4-Alice-v1.2-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Agares-31B-v1I1-Q4_K_M30.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma4-Gutenberg-31B-HereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Gemsicle-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Gembrain-31B-it-uncensored-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Melinoe-Gemma4-31B-VL-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
G4-MeroMero-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Glistening-Gem-31B-v1.0I1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Melinoe-Gemma4-31B-VLI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-31B-Storymaxxed3I1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Huihui-gemma-4-31B-it-qat-q4_0-unquantized-abliteratedI1-Q4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-Queen-it-qat-q4_0-unquantizedI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-qat-q4_0-unquantized-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-AssGuard-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
copywriter-gemma4-31bI1-Q4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-heretic-finetuneI1-Q4_K_M30.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Garnet-V2-31B-it-ultra-uncensored-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-Claude-Opus-Distill-v2Q4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-abliterated-v3I1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Harmonia-31B-uncensored-hereticQ4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-noloopI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Webs-Sejong-31B-v7I1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Lilith-31B-v1.0I1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
JGOS-31B-ThinkI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-MergemaxxedI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
K1-v6-zeroI1-Q4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-uncensored-hereticQ4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Queen-31B-it-uncensored-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Sphinsikus-Chronist-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-hereticI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma4-31B-Finetuned-V2I1-Q4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-31B-storymaxxedI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-31B-Fable-5-Agent-DistillQ4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-uncensoredQ4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-31B-storymaxxed2I1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Giftige-Blume-31B-v2Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-Grand-Horror-X-INTENSE-HERETIC-UNCENSORED-ThinkingI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-Mystery-Fine-Tune-HERETIC-UNCENSORED-ThinkingI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-ThinkingI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Huihui-gemma-4-31B-it-abliterated-v2I1-Q4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Gemma-4-Queen-31B-itI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-abliteratedI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31b-it-heretic-araI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
Monika-31BI1-Q4_K_M31.3B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
gemma-4-31B-it-uncensoredQ4_K_M32.7B17.40 GiB11.25 GiB29.73 GiB0.03 GiB18±22%
●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 Tesla V100 32GB run?
1859 of 2118 indexed open-weight models fit a Tesla V100 32GB at 131,072 context with q8_0 KV cache, the largest being Qwen3-VL-32B-Thinking at Q2_K_L. That covers text, vision-language, image, video and speech models.
How much usable memory does a Tesla V100 32GB actually have?
Its nameplate is 32 GB, but about 29.76 GiB is available to a model once driver and compositor overhead is accounted for.
Is a Tesla V100 32GB fast for local AI?
Its memory bandwidth is 900 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.