AI models
DeepSpeech2 (English)
Training compute
2.6×10¹⁹ FLOP
Parameters
38M
Published
Dec 8, 2015
DeepSpeech2 (English) is an AI model developed by Baidu Research - Silicon Valley AI Lab (United States), first published in December 2015. It works in the speech domain, on tasks such as speech recognition (asr). It counts among the frontier models: the systems trained with the most compute of their moment.
Training it took an estimated 2.6×10¹⁹ FLOP of compute (estimation method: operation counting,third-party estimation). The model has 38,000,000 parameters. It was trained on roughly 716.4M datapoints. Training ran on 16 NVIDIA GeForce GTX TITAN X for about 120 hours. The compute alone is estimated at $214 in 2023 dollars.
The reference paper has 3,150 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- Baidu Research - Silicon Valley AI Lab
- Country of organization
- United States
- Domain
- Speech
- Task
- Speech recognition (ASR)
- Training compute
- 2.6×10¹⁹ FLOP
- Compute estimation method
- Operation counting, Third-party estimation
- Parameters
- 38,000,000
- Dataset size
- 716.4M
- Training hardware
- NVIDIA GeForce GTX TITAN X
- Chips used
- 16
- Training time
- 120 h
- Chip-hours
- 1.9K
- Training power draw
- 8.5 kW
- Training cost (2023 USD)
- $214
- Numerical format
- FP32
- Citations
- 3,150
- Epoch confidence
- Confident
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
← All ai models