AI models
KataGo
Training compute
2.3×10¹⁹ FLOP
Parameters
2.5M
Published
Feb 27, 2019
KataGo is an AI model developed by Jane Street (United States), first published in February 2019. It works in the games domain, on tasks such as go.
Training it took an estimated 2.3×10¹⁹ FLOP of compute (estimation method: hardware). The model has 2,500,000 parameters. It was trained on roughly 241M datapoints. Training ran on NVIDIA Tesla V100 DGXS 16 GB for about 456 hours. The compute alone is estimated at $105 in 2023 dollars.
Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 111 citations. Epoch AI rates the confidence of this record as speculative.
Full record
- Organization
- Jane Street
- Country of organization
- United States
- Domain
- Games
- Task
- Go
- Training compute
- 2.3×10¹⁹ FLOP
- Compute estimation method
- Hardware
- Parameters
- 2,500,000
- Dataset size
- 241M
- Training hardware
- NVIDIA Tesla V100 DGXS 16 GB
- Training time
- 456 h
- Training cost (2023 USD)
- $105
- Numerical format
- FP16
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Citations
- 111
- Epoch confidence
- Speculative
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