Big-Little Net (speech)

Training compute

4.3×10¹⁷ FLOP

Parameters

3.3M

Published

Jul 10, 2018

Big-Little Net (speech) is an AI model developed by IBM (United States), first published in July 2018. It works in the speech domain, on tasks such as speech recognition (asr).

Training it took an estimated 4.3×10¹⁷ FLOP of compute (estimation method: operation counting). The model has 3,320,000 parameters. It was trained on roughly 720M datapoints.

Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 101 citations. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: IBM
Country of organization: United States
Domain: Speech
Task: Speech recognition (ASR)
Training compute: 4.3×10¹⁷ FLOP
Compute estimation method: Operation counting
Parameters: 3,320,000
Dataset size: 720M
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Citations: 101
Epoch confidence: Speculative