NVIDIA · datacenter

Tesla V100 16GB

Tesla V100 16GB has 16 GB of VRAM at 900 GB/s — about 14.88 GiB usable after driver and compositor overhead. 1862 of 2118 indexed models fit at 4K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
16 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
text 1597vision language 162audio asr 39video 15embedding 26audio tts 21image 2

What fits at 4K context

largest quantization that fits, per model · 1862 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Le-Chaton-Slim-23BMoEI1-Q4_123.3B13.65 GiB0.22 GiB14.88 GiB0.00 GiB78±37%
Qwen3.5-27B-Engineer-Deckard-GeminiI1-IQ4_XS27.7B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.5-27B-HERETIC-Polaris-Advanced-Thinking-Alpha-uncensoredI1-IQ4_XS27.4B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.5-27B-Deckard-PKD-Heretic-Uncensored-ThinkingI1-IQ4_XS27.4B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Huihui-Qwen3.5-27B-abliteratedI1-IQ4_XS27.8B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.5-27B-Unredacted-MAXI1-IQ4_XS27.4B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.5-27B-hereticI1-IQ4_XS27.4B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.5-27B-DerestrictedI1-IQ4_XS27.8B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.5-27B-Claude-4.6-Opus-Reasoning-DistilledI1-IQ4_XS27.8B13.68 GiB0.13 GiB14.88 GiB0.00 GiB37±22%
Qwen3.6-35B-A3B-REAM-160-ru-agentMoEQ4_123.6B13.82 GiB0.04 GiB14.87 GiB0.01 GiB174±37%
granite-20b-code-instruct-8kQ5_K_M20.1B13.79 GiB0.00 GiB14.87 GiB0.01 GiB37±22%
granite-20b-code-base-8kI1-Q5_K_M20.1B13.79 GiB0.00 GiB14.87 GiB0.01 GiB37±22%
granite-34b-code-base-8kI1-IQ3_S33.7B13.79 GiB0.00 GiB14.87 GiB0.01 GiB37±22%
Skywork-R1V3-38BIQ3_M38.4B13.79 GiB0.00 GiB14.86 GiB0.02 GiB37±22%
Qwen3-Coder-REAP-25B-A3BMoEQ4_K_S24.9B13.66 GiB0.20 GiB14.85 GiB0.03 GiB135±37%
Gemma-4-Gembrain-X-Core-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Gembrain-X-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-31B-Isometry-Fabled-PersonaI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Versipellis-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma4-Gutenberg-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
G4-MeroMero-31B-uncensored-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Novelist-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Wanabi-Gemma4-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
G4-Alice-v1.2-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Agares-31B-v1I1-IQ3_S30.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma4-Gutenberg-31B-HereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Gemsicle-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Gembrain-31B-it-uncensored-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Melinoe-Gemma4-31B-VL-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
G4-MeroMero-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Glistening-Gem-31B-v1.0I1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Melinoe-Gemma4-31B-VLI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-31B-Storymaxxed3I1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Huihui-gemma-4-31B-it-qat-q4_0-unquantized-abliteratedI1-IQ3_S32.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-Queen-it-qat-q4_0-unquantizedI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-qat-q4_0-unquantized-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-AssGuard-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
copywriter-gemma4-31bI1-IQ3_S32.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-heretic-finetuneI1-IQ3_S30.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Garnet-V2-31B-it-ultra-uncensored-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-Claude-Opus-Distill-v2Q3_K_S32.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-abliterated-v3I1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Harmonia-31B-uncensored-hereticQ3_K_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-noloopI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Webs-Sejong-31B-v7I1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Lilith-31B-v1.0I1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
JGOS-31B-ThinkI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-MergemaxxedI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
K1-v6-zeroI1-IQ3_S32.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Queen-31B-it-uncensored-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-Sphinsikus-Chronist-31BI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-hereticI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma4-31B-Finetuned-V2I1-IQ3_S32.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-31B-storymaxxedI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-31B-Fable-5-Agent-DistillQ3_K_S32.7B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
Gemma-4-31B-storymaxxed2I1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-Grand-Horror-X-INTENSE-HERETIC-UNCENSORED-ThinkingI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-Mystery-Fine-Tune-HERETIC-UNCENSORED-ThinkingI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±22%
gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-ThinkingI1-IQ3_S31.3B12.82 GiB0.95 GiB14.85 GiB0.03 GiB37±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 16GB run?
1862 of 2118 indexed open-weight models fit a Tesla V100 16GB at 4,096 context with q8_0 KV cache, the largest being Le-Chaton-Slim-23B at I1-Q4_1. That covers text, vision-language, image, video and speech models.
How much usable memory does a Tesla V100 16GB 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 Tesla V100 16GB 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.