BART
Denoising sequence-to-sequence pre-training for NLP
A denoising sequence-to-sequence pre-trained language model family introduced by Facebook AI Research for natural language generation, translation, and comprehension.
Pricing
Official site · Jul 1, 2026, 7:26 AM
Pricing notes were collected, but there are no normalized numeric fields to display yet.
No public pricing is listed on the official Meta BART release page or fairseq example documentation; no subscription, usage, or per-token price is stated.
Meta’s official BART release materials and fairseq example docs do not list any public commercial pricing or API fees for BART. It appears to be distributed as an open model rather than a priced service, so no normalized price could be recorded.
Model Intelligence
Context window
1,024 tokens
Benchmarkable
No
Model level
family
Recent stories
0 linked stories
No linked stories yet.