NVIDIA · datacenter

Tesla P100 16GB

Tesla P100 16GB has 16 GB of VRAM at 732 GB/s — about 14.88 GiB usable after driver and compositor overhead. 1862 of 2118 indexed models fit at 8K context with q4_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
16 GB
HBM2
Bandwidth
732 GB/s
4096-bit bus
Tensor FP16
—
dense
TDP
250 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 8K context

largest quantization that fits, per model · 1862 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
gpt-oss-20b-hereticMoEQ4_K_S20.9B13.83 GiB0.06 GiB14.88 GiB0.00 GiB91±37%
Qwen3.6-35B-A3B-REAM-160-ru-agentMoEQ4_123.6B13.82 GiB0.04 GiB14.87 GiB0.01 GiB144±37%
Yi-34B-200K-DARE-megamerge-v8I1-IQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
dolphin-2.9.1-yi-1.5-34bI1-IQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
OrionStar-Yi-34B-Chat-LlamaI1-IQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
Yi-34B-200K-LlamafiedI1-IQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
Yi-1.5-34BIQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
Merged-RP-Stew-V2-34BI1-IQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
Capybara-Tess-Yi-34B-200KI1-IQ3_XS34.4B13.26 GiB0.53 GiB14.87 GiB0.01 GiB30±22%
granite-20b-code-instruct-8kQ5_K_M20.1B13.79 GiB0.00 GiB14.87 GiB0.01 GiB30±22%
granite-20b-code-base-8kI1-Q5_K_M20.1B13.79 GiB0.00 GiB14.87 GiB0.01 GiB30±22%
granite-34b-code-base-8kI1-IQ3_S33.7B13.79 GiB0.00 GiB14.87 GiB0.01 GiB30±22%
North-Mini-Code-1.0MoEQ3_K_L30.5B13.74 GiB0.15 GiB14.87 GiB0.01 GiB125±37%
EXAONE-4.0-32BIQ3_M32.0B13.39 GiB0.38 GiB14.87 GiB0.01 GiB30±22%
Qwen3-Coder-REAP-25B-A3BMoEQ4_K_S24.9B13.66 GiB0.21 GiB14.87 GiB0.01 GiB111±37%
EuroLLM-22B-Instruct-2512Q4_122.6B13.32 GiB0.47 GiB14.86 GiB0.02 GiB30±22%
Skywork-R1V3-38BIQ3_M38.4B13.79 GiB0.00 GiB14.86 GiB0.02 GiB30±22%
Nous-Capybara-limarpv3-34BI1-IQ3_XXS34.4B13.24 GiB0.53 GiB14.85 GiB0.03 GiB30±22%
gemma-4-26B-A4B-itMoEIQ4_NL26.5B13.69 GiB0.17 GiB14.85 GiB0.03 GiB30±22%
Fallen-Gemma3-27B-v1Q3_K_L27.4B13.54 GiB0.26 GiB14.84 GiB0.04 GiB30±22%
Qwen3.6-28BMoEI1-Q3_K_L28.2B13.77 GiB0.04 GiB14.82 GiB0.06 GiB158±37%
Qwen3.5-28BMoEI1-Q3_K_L28.7B13.77 GiB0.04 GiB14.82 GiB0.06 GiB158±37%
Delphi-25B-SimpleRL-MathI1-Q3_K_M25.0B11.39 GiB2.35 GiB14.82 GiB0.06 GiB30±22%
dolphin-2.9.3-mistral-7B-32kF167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Mistral-7B-Instruct-v0.3-ParasiteF167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Mistral-7B-Instruct-v0.3-JbliteratedF167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Mistral-7B-Instruct-v0.3F167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Mistral-7B-v0.3F167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Mistral-7B-v0.3-Chinese-ChatF167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
mistral-7b-v0.3-bnb-4bitBF167.5B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Mathstral-7B-v0.1F167.2B13.50 GiB0.28 GiB14.82 GiB0.06 GiB30±22%
Devstral-Small-2-24B-Instruct-2512Q4_K_M24.0B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Voxtral-Small-24B-2507Q4_K_M24.3B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Magistral-Small-2509Q4_K_M24.0B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Magistral-Small-2507Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Transformed-Journey-24BI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Magistry-24B-v1.1I1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Mergedonia-AETHER-24B-v1aI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Mergedonia-AETHER-24B-v1bI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Slimaki-Tavern-24B-v1.3I1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Maginum-Cydoms-24BI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Maginum-Cydoms-24B-absolute-heresyI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Dolphin3.0-R1-Mistral-24BQ4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Dolphin3.0-Mistral-24BQ4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Mistral-Small-3.2-24B-Instruct-2506-ultra-uncensored-hereticI1-Q4_K_M24.0B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Huihui-Mistral-Small-3.2-24B-Instruct-2506-abliterated-llamacppfixedI1-Q4_K_M24.0B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Dans-PersonalityEngine-V1.2.0-24bI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Mistral-Small-3_2-24B-Instruct-2506-antislop.v2I1-Q4_K_M24.0B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Cydonia_VistralQ4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Mistral-Small-3.2-24B-Instruct-2506Q4_K_M24.0B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Dans-PersonalityEngine-V1.3.0-24bI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Devstral-Small-2507Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Goetia-24B-v1.1I1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Devstral-Small-2505Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
MS3.2-PaintedFantasy-v3-24BI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
RP-Spectrum-24BI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
MS3.2-PaintedFantasy-v4.1-24B-ultra-uncensored-heretic-v2I1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Magidonia-24B-v4.3-heretic-v1.2I1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
Magidonia-24B-v4.3-absolute-heresyI1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±22%
MagiSeek-Pro-V1I1-Q4_K_M23.6B13.35 GiB0.35 GiB14.82 GiB0.06 GiB30±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 P100 16GB run?
1862 of 2118 indexed open-weight models fit a Tesla P100 16GB at 8,192 context with q4_0 KV cache, the largest being gpt-oss-20b-heretic at Q4_K_S. That covers text, vision-language, image, video and speech models.
How much usable memory does a Tesla P100 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 P100 16GB fast for local AI?
Its memory bandwidth is 732 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.