Amazon Titan
Amazon Titan is an AI model developed by Amazon (United States), first published in September 2023. It works in the language and image generation domain, on tasks such as semantic search, image generation, language modeling/generation and 4 more. It counts among the frontier models: the systems trained with the most compute of their moment.
Training it took an estimated 4.8×10²⁴ FLOP of compute (estimation method: hardware,operation counting). The model has 200,000,000,000 parameters. It was trained on roughly 4T datapoints. Training ran on 13,760 NVIDIA A100 for about 1.2K hours. The compute alone is estimated at $8M in 2023 dollars.
Access: API access. Its weights are not openly released. Epoch AI rates the confidence of this record as likely.
- Organization
- Amazon
- Country of organization
- United States
- Domain
- Language, Image generation
- Task
- Semantic search, Image generation, Language modeling/generation, Code generation, Chat, Text-to-image, Translation
- Training compute
- 4.8×10²⁴ FLOP
- Compute estimation method
- Hardware, Operation counting
- Parameters
- 200,000,000,000
- Dataset size
- 4T
- Training hardware
- NVIDIA A100
- Chips used
- 13,760
- Training time
- 1,152 h
- Training power draw
- 10.9 MW
- Training cost (2023 USD)
- $8M
- Model accessibility
- API access
- Open weights
- No
- Epoch confidence
- Likely