How the ranking works. Each configuration was asked for four scenes. It earns one point for every scene where at least 3 of the 4 reviewers agreed the requested event happened, and a second point only where they also agreed the source or cause was right — a Factual Score out of 8. Ties go to more scenes with every factual question right; anything still equal shares its rank. A refused scene is still one of the four: it earns nothing and is marked R, and it is coverage, not a bad video. A scene without a 3-of-4 consensus earns nothing and is never counted as a failure. Sound, timing and people ratings are shown beside the score and never break a tie.

Overall Ranking V1 was adopted after the Core R01 reveal; it was not preregistered for R01. Ranking fartbench-core-overall-ranking-v1 · RANKING_V1.json (SHA-256 c6c354076bc6…) · every underlying number is on each configuration's page.

Official Core R01 leaderboard, Overall Ranking V1
Rank Configuration Factual scoreScore Event Right source / causeCause Every question Sound Timing People Reviewers Coverage
1= Seedance 2.5 ByteDance Seedance 8/8 4/4 4/4 4/4 3.9 n=16 · 0 N/A 4.2 n=16 · 0 N/A 3.8 n=16 · 0 N/A 4/clip 4/4 generated
1= Vidu Omni Pro 3.2a Vidu 8/8 4/4 4/4 4/4 4.1 n=16 · 0 N/A 4.3 n=16 · 0 N/A 4.0 n=16 · 0 N/A 4/clip 4/4 generated
3= Wan 3 Alibaba Wan 3 6/8 4/4 2/4 2/4 3.2 n=16 · 0 N/A 3.4 n=16 · 0 N/A 3.6 n=16 · 0 N/A 4/clip 4/4 generated
3= FLUX 3 Draft Black Forest Labs FLUX 3 6/8 4/4 2/4 2/4 2.4 n=15 · 1 N/A 3.1 n=15 · 1 N/A 3.1 n=16 · 0 N/A 4/clip 4/4 generated
3= MiniMax H3 768p local MiniMax H3 6/8 4/4 2/4 2/4 2.7 n=15 · 1 N/A 3.0 n=15 · 1 N/A 3.1 n=16 · 0 N/A 4/clip 4/4 generated
6 FLUX 3 Full Black Forest Labs FLUX 3 5/8 4/4 1/4 0/4 3.1 n=15 · 1 N/A 3.1 n=15 · 1 N/A 3.1 n=16 · 0 N/A 4/clip 4/4 generated
7 Veo 3.1 Standard Google Veo 3.1 4/8 3/4 1/4 1/4 3.2 n=16 · 0 N/A 2.6 n=16 · 0 N/A 3.1 n=16 · 0 N/A 4/clip 4/4 generated
8= HappyHorse 1.1 HappyHorse 3/8 2/4 1/4 1/4 1.6 n=10 · 6 N/A 3.3 n=11 · 5 N/A 3.5 n=12 · 4 N/A 4/clip 4/4 generated
8= Sora 2 Pro OpenAI Sora 3/8 2/4 1/4 1/4 3.1 n=10 · 6 N/A 3.3 n=10 · 6 N/A 2.9 n=12 · 4 N/A 4/clip 4/4 generated
10= Kling VIDEO 3.0 Omni Kling 3 2/8 1/4 R 1/4 1/4 3.0 n=4 · 8 N/A 3.5 n=6 · 6 N/A 3.1 n=9 · 3 N/A 4/clip 3/4 generated 1 refused
10= MiniMax H3 2K API MiniMax H3 2/8 1/4 1/4 1/4 2.4 n=7 · 9 N/A 2.9 n=8 · 8 N/A 3.0 n=11 · 5 N/A 4/clip 4/4 generated
10= Sora 2 OpenAI Sora 2/8 1/4 1/4 1/4 2.3 n=11 · 5 N/A 2.9 n=11 · 5 N/A 2.8 n=11 · 5 N/A 4/clip 4/4 generated
13= Veo 3.1 Fast Google Veo 3.1 1/8 1/4 0/4 0/4 2.3 n=16 · 0 N/A 1.7 n=14 · 2 N/A 2.4 n=16 · 0 N/A 4/clip 4/4 generated
13= LTX-2.5 Fast LTX-2.5 1/8 1/4 0/4 0/4 1.0 n=6 · 10 N/A 2.8 n=6 · 10 N/A 2.7 n=10 · 6 N/A 4/clip 4/4 generated
15= LTX-2.5 Pro LTX-2.5 0/8 0/4 0/4 0/4 1.0 n=2 · 14 N/A 3.5 n=2 · 14 N/A 2.9 n=11 · 5 N/A 4/clip 4/4 generated
15= Grok Imagine Video 1.5 xAI Grok Imagine 0/8 0/4 0/4 0/4 1.0 n=1 · 15 N/A 3.0 n=1 · 15 N/A 3.4 n=9 · 7 N/A 4/clip 4/4 generated

= shared rank · R the provider refused that scene (coverage, not quality) · n = numeric ratings and N/A answers set aside; fewer than 8 shown grey. The frozen reviewer evidence behind every row is on each configuration's page and on the full results.