DeepSeek · open weights
DeepSeek V4.1-Flash
DeepSeek's new default model and the replacement for the retired V4-Flash: MIT weights, 1M context, 384K max output, and image input.
- Organisation
- DeepSeek
- Weights published
- Licence
- MIT
- Size
- 552B total MoE (encoder-decoder; 8B active for input, 16B for output)
- Modalities
- textimage
- Comparison catalog
- DeepSeek V4.1 Flash profile
Practical fit
The low-cost hosted route ($0.30/$1.20 peak, $0.15/$0.60 off-peak) with an open-weight exit for teams that can serve a 552B MoE.
Deployment note
Hosted API id is `deepseek-flash`; the legacy `deepseek-v4-flash` ids temporarily route here. Self-hosting still means serving the full expert set.