AI models
VILA-13B
Training compute
2.3×10²¹ FLOP
Parameters
13.4B
Published
Dec 12, 2023
VILA-13B is an AI model developed by NVIDIA and Massachusetts Institute of Technology (MIT) (United States), first published in December 2023. It works in the multimodal, language and vision domain, on tasks such as chat, visual question answering, image captioning and 2 more.
Training it took an estimated 2.3×10²¹ FLOP of compute. The model has 13,350,839,296 parameters. It was trained on roughly 32.4B datapoints. Training ran on 128 NVIDIA A100 SXM4 80 GB.
Access: Open weights (non-commercial). Its weights are openly available. It is built on top of Llama 2-13B,CLIP (ViT L/14@336px). The reference paper has 827 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- NVIDIA, Massachusetts Institute of Technology (MIT)
- Country of organization
- United States
- Domain
- Multimodal, Language, Vision
- Task
- Chat, Visual question answering, Image captioning, Language modeling/generation, Question answering
- Training compute
- 2.3×10²¹ FLOP
- Parameters
- 13,350,839,296
- Dataset size
- 32.4B
- Training hardware
- NVIDIA A100 SXM4 80 GB
- Chips used
- 128
- Training power draw
- 101.5 kW
- Numerical format
- BF16
- Model accessibility
- Open weights (non-commercial)
- Open weights
- Yes
- Base model
- Llama 2-13B, CLIP (ViT L/14@336px)
- Citations
- 827
- Epoch confidence
- Confident
More from NVIDIA,Massachusetts Institute of Technology (MIT)
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