GPUs & Macs › NVIDIA Tesla P100

NVIDIA Tesla P100: local LLM speed

Llama 2 7B (Q4_0) generates 66.5 tokens/s on the NVIDIA Tesla P100 (Vulkan), and processes prompts at 686 tokens/s.

Measured results

ModelGenerate
tok/s		Prompt
tok/s		QuantBackendTestSource
Llama 2 7B66.5686Q4_0Vulkan FA ontg128llama.cpp #10879 details
Llama 2 7B63.2678Q4_0Vulkan FA offtg128llama.cpp #10879 details
Llama 2 7B62.0787Q4_0CUDA FA ontg128llama.cpp #15013 details
Llama 2 7B58.4761Q4_0CUDA FA offtg128llama.cpp #15013 details
Llama 3.2 1B Instruct1202,380Q4_K_Mllamafile LocalScore avg · n=2LocalScore LocalScore details
Llama 3.1 8B Instruct32.3399Q4_K_Mllamafile LocalScore avg · n=2LocalScore LocalScore details
Qwen2.5 14B Instruct19.8193Q4_K_Mllamafile LocalScore avg · n=2LocalScore LocalScore details

What fits on the NVIDIA Tesla P100

Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 66.5 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BUD-IQ3_XXS10.9 GB23.2 estimate
gemma-4-12B-itQ8_012.7 GB20.1 estimate
Qwen3.5-9BQ8_09.5 GB26.7 estimate
Qwen3.5-4BQ8_04.5 GB56.7 estimate
Qwen3.6-35B-A3BUD-IQ3_XXS13.2 GB112 estimate
gemma-4-26B-A4B-itUD-IQ4_XS13.6 GB60.8 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB26.6 estimate
gemma-4-E4B-itQ8_08.2 GB31.0 estimate
Muse-Glimmer-30BUD-IQ3_XXS13.1 GB19.4 estimate
Qwen3-4BQ8_04.3 GB59.4 estimate
gemma-4-31B-itUD-IQ3_XXS11.8 GB21.5 estimate
gemma-4-12B-it-qat-q4_0-unquantizedUD-Q4_K_XL6.7 GB37.8 estimate
gemma-3-1b-itQ8_01.1 GB212 estimate
gemma-4-E2B-itQ8_05.0 GB51.2 estimate

Open in the picker

Similar hardware

NVIDIA Tesla T4 · NVIDIA Tesla V100 16GB · NVIDIA Tesla K80 · NVIDIA A30 · NVIDIA Tesla P40 · NVIDIA Tesla M40 · NVIDIA Tesla P4 · NVIDIA L4