Model leaderboard

Every model we’ve run through the standard Glif tests, scored out of 10 by an LLM judge. Overall needs 3 judged tests to rank; pick a capability to rank on it alone.

Identity preservation

Faces/characters/objects from the input remain recognizably the same in the output (likeness, clothing, distinguishing features).

1
Image to Video (Kling O3 Pro)
1 test
8.0
2
Image to Video (Kling 2.5 Pro)
1 test
8.0
3
Seedance Pro 1.5
1 test
8.0
4
Image to Video (Vidu Q3)
3 tests
7.8
5
Reference to Video (Grok Imagine)
1 test
7.5
6
Image to Video (Veo 3.1 Fast)
1 test
7.5
7
Image to Video (Veo 3.1)
1 test
7.5
8
GPT Image 1.5
6 tests
7.3
9
Grok Imagine Video
1 test
7.0
10
Image to Video (Hailuo 2.3)
1 test
7.0
11
Image to Video (Kling 2.6 Pro)
1 test
7.0
12
Motion Control (Kling 2.6 Pro)
2 tests
7.0
13
Reference to Video (Kling O3 Pro)
1 test
7.0
14
Image to Video (Kling V3 Pro)
1 test
7.0
15
Image to Video (Kling V3 Standard)
1 test
7.0
16
FLUX Kontext Image Editor
4 tests
6.1
17
Qwen Image Generator with LoRA
4 tests
5.5
18
Edit Video (Grok Imagine)
1 test
2.0