#1▼$0.30/sByteDanceSeedance 2.5
#2▲$0.03/sAlibabaWan 3.0
#3▼$0.12/sByteDanceSeedance 2.0
#4▲$0.04/sPruna AIP-Video 2 Pro
#5▼$0.15/sBlack Forest LabsFLUX 3

Swipe for more top models

V-Benchmark changelog

← Evaluation Changelogs

V-Benchmark 2.3.0

Methodology updates

  • Prompt suite updated
  • Rubric updated across 2 categories. Prompt Adherence: 15 checklist items retired; Prompt Adherence: 5 checklist items revised; Scene Consistency: 4 checklist items retired; Scene Consistency: 12 checklist items revised. Scores recalculated from retained answers.
  • Quality + Reliability replaces delivered-file scoring and now accounts for 2.5% of the Megaton Index. Motion compensation was corrected and the published models were remeasured. No Black Frames and No Flash/Strobe remain diagnostic checks without score weight. Missing measurements no longer count as passing or failing, and automated technical scores no longer require manual review.
  • Weighting updated. Quality + Reliability: 1 checklist item reweighted.

V-Benchmark 2.2.0

Methodology updates

  • Rubric updated across 2 categories. Prompt Adherence: 30 checklist items retired; Scene Consistency: 9 checklist items retired. Scores recalculated from retained answers.
  • Weighting updated across 3 categories. Prompt Adherence: 1 checklist item reweighted; Scene Consistency: 1 checklist item reweighted; Object & Product Fidelity: 1 checklist item reweighted.

V-Benchmark 2.1.0

Methodology updates

  • Rubric updated. Prompt Adherence: 1 checklist item retired. Scores recalculated from retained answers.
  • Weighting updated across 2 categories. Prompt Adherence: 1 checklist item reweighted; Object & Product Fidelity: 1 checklist item reweighted.

V-Benchmark 2.0.0

Methodology updates

  • Violation counts published

V-Benchmark 2.0.0

Methodology updates

  • Methodology published