Skip to content

GPUs Rent a GPU Train New Inference New Models Model benchmark API & MCP

Tools

CompareUp to 4 GPUs side by side CloudsEvery provider and its prices Model viewerOpen an AI model in the browser

Compare Clouds Model viewer

Prices checked 13:04 UTC

Loading…

How GridPup works Training Model benchmark Clouds Contact Privacy Terms API

Prices come straight from each cloud and update every 15 minutes. Prices are per GPU per hour in US dollars; a month is 720 hours.

© GridPup

whisper-small: GPU requirements and price to run | GridPup

Models / whisper-small

onnx-community/whisper-small

Speech to text · served with ONNX Runtime · ↓ 134.6K downloads a month · View on Hugging Face

Weights: 610 MB
GPU memory needed: about 1.9 GB
Memory · disk: 8 GB · 30 GB
Licence: not stated

Versions

Smaller versions (8-bit, 4-bit) need less GPU memory and answer a little less precisely.

full precisiononnx/decoder_model.onnx610 MB · 1.9 GB GPU 4-bit (bnb)onnx/decoder_model_bnb4.onnx230 MB · 1.3 GB GPU float16onnx/decoder_model_fp16.onnx310 MB · 1.5 GB GPU 8-bitonnx/decoder_model_int8.onnx160 MB · 1.2 GB GPU full precisiononnx/decoder_model_merged.onnx620 MB · 1.9 GB GPU 4-bit (bnb)onnx/decoder_model_merged_bnb4.onnx230 MB · 1.3 GB GPU float16onnx/decoder_model_merged_fp16.onnx310 MB · 1.5 GB GPU 8-bitonnx/decoder_model_merged_int8.onnx160 MB · 1.2 GB GPU 4-bitonnx/decoder_model_merged_q4.onnx230 MB · 1.3 GB GPU 8-bitonnx/decoder_model_merged_quantized.onnx160 MB · 1.2 GB GPU 8-bitonnx/decoder_model_merged_uint8.onnx160 MB · 1.2 GB GPU 4-bitonnx/decoder_model_q4.onnx230 MB · 1.3 GB GPU 8-bitonnx/decoder_model_quantized.onnx160 MB · 1.2 GB GPU 8-bitonnx/decoder_model_uint8.onnx160 MB · 1.2 GB GPU full precisiononnx/decoder_with_past_model.onnx560 MB · 1.8 GB GPU 4-bit (bnb)onnx/decoder_with_past_model_bnb4.onnx220 MB · 1.3 GB GPU float16onnx/decoder_with_past_model_fp16.onnx280 MB · 1.4 GB GPU 8-bitonnx/decoder_with_past_model_int8.onnx140 MB · 1.2 GB GPU 4-bitonnx/decoder_with_past_model_q4.onnx220 MB · 1.3 GB GPU 8-bitonnx/decoder_with_past_model_quantized.onnx140 MB · 1.2 GB GPU 8-bitonnx/decoder_with_past_model_uint8.onnx140 MB · 1.2 GB GPU full precisiononnx/encoder_model.onnx350 MB · 1.5 GB GPU 4-bit (bnb)onnx/encoder_model_bnb4.onnx60 MB · 1.1 GB GPU float16onnx/encoder_model_fp16.onnx180 MB · 1.3 GB GPU 8-bitonnx/encoder_model_int8.onnx90 MB · 1.1 GB GPU 4-bitonnx/encoder_model_q4.onnx70 MB · 1.1 GB GPU 8-bitonnx/encoder_model_quantized.onnx90 MB · 1.1 GB GPU 8-bitonnx/encoder_model_uint8.onnx90 MB · 1.1 GB GPU

Hardware and deploy

Deploying ONNX models from the catalog is coming soon. Today, download the files from Hugging Face and upload them in Inference.

Prices are per hour for the whole machine and its disk, right now. Speed is Vast's DLPerf score of the machine, relative to the cheapest. Best value: the most speed per dollar. Only machines with a Vast reliability above 95% that download at 400 Mbit/s or more are offered: the machine fetches the model itself.