Does increasing Dory's loop count make it a larger model?
The published parameter count remains 4 billion because the recurrent layers are shared. More loops mean more sequential computation per inference, so parameter count alone does not describe the full runtime cost or latency.
Answered in
NVIDIA Dory Makes Inference Compute a Knob. What Does That Buy?NVIDIA Dory 1.0 reuses transformer layers at inference. Its adjustable loop count trades more computation for task performance, not free capability.
Read the full analysisOther questions this article answers
More how ai actually works questions
- What does Google's Deep Review accept?
- Did the Paper Assistant Tool replace conference peer review?
- Are uploaded documents used to train Google models?
- When does Google Workspace begin enforcing the MCP OAuth client marker?
- Will this change break regular Google Workspace REST API integrations?
- What should a team do if one OAuth client serves agents and regular integrations?
- What is NVIDIA Dory 1.0?
- Are the model card's benchmark numbers independent results?
Every answer on Crashtech is written by the editor of the article it comes from — never auto-summarised. Browse all answers or the How AI Actually Works beat.