AI models
GPT-SW3
Training compute
1.7×10²¹ FLOP
Parameters
3.5B
Published
Jun 25, 2022
GPT-SW3 is an AI model developed by AI Sweden and RISE (Sweden), first published in June 2022. It works in the language domain, on tasks such as language modeling.
Training it took an estimated 1.7×10²¹ FLOP of compute (estimation method: hardware,operation counting). The model has 3,500,000,000 parameters. It was trained on roughly 101.7B datapoints. Training ran on NVIDIA A100 for about 60 hours.
Access: Unreleased. Its weights are not openly released. The reference paper has 20 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- AI Sweden, RISE
- Country of organization
- Sweden
- Domain
- Language
- Task
- Language modeling
- Training compute
- 1.7×10²¹ FLOP
- Compute estimation method
- Hardware, Operation counting
- Parameters
- 3,500,000,000
- Dataset size
- 101.7B
- Training hardware
- NVIDIA A100
- Training time
- 60 h
- Model accessibility
- Unreleased
- Open weights
- No
- Citations
- 20
- 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