AI models
DreamLLM
Training compute
7.5×10²⁰ FLOP
Parameters
7B
Published
Sep 20, 2023
DreamLLM is an AI model developed by Xi’an Jiaotong University, Megvii Inc, Tsinghua University and Huazhong University of Science and Technology (China), first published in September 2023. It works in the multimodal, language, vision and image generation domain, on tasks such as language modeling/generation, vision-language generation and image generation.
Training it took an estimated 7.5×10²⁰ FLOP of compute (estimation method: hardware). The model has 7,000,000,000 parameters. It was trained on roughly 70.4B datapoints. Training ran on 128 NVIDIA A800 PCIe 40 GB for about 17.5 hours.
The reference paper has 318 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- Xi’an Jiaotong University, Megvii Inc, Tsinghua University, Huazhong University of Science and Technology
- Country of organization
- China
- Domain
- Multimodal, Language, Vision, Image generation
- Task
- Language modeling/generation, Vision-language generation, Image generation
- Training compute
- 7.5×10²⁰ FLOP
- Compute estimation method
- Hardware
- Parameters
- 7,000,000,000
- Dataset size
- 70.4B
- Training hardware
- NVIDIA A800 PCIe 40 GB
- Chips used
- 128
- Training time
- 18 h
- Chip-hours
- 2.2K
- Training power draw
- 63.6 kW
- Citations
- 318
- 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