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.

Photorealism

Output is indistinguishable from a real photograph: plausible lighting, optics, materials, and imperfections. No 'AI sheen', waxy skin, or over-smoothed textures.

1
FLUX.1 Dev Image Generator
2 tests
7.8
2
FLUX.2 Flex Image Generator
2 tests
7.5
3
FLUX.2 Pro Image Generator
2 tests
7.5
4
Nano Banana Pro Image Generator
1 test
7.5
5
GPT Image 1.5
6 tests
7.3
6
FLUX.2 Klein 9B Image Generator
2 tests
7.0
7
FLUX.2 Max Image Generator
2 tests
7.0
8
FLUX.2 Turbo Image Generator
1 test
7.0
9
Image to Video (Kling V3 4K)
1 test
7.0
10
Seedream V4.5
1 test
7.0
11
Video (Veo 3.1 Lite)
1 test
7.0
12
FLUX.1 Schnell Image Generator
2 tests
6.5
13
Qwen Image Generator with LoRA
4 tests
6.5
14
FLUX Kontext Image Editor
2 tests
6.3
15
Grok Imagine Video
1 test
5.0
16
Image to Video (Veo 3.1 Fast)
1 test
5.0
17
Image to Video (Veo 3.1)
1 test
5.0
18
Image to Video (Vidu Q3)
1 test
5.0
19
Image to Video (Kling O3 Pro)
1 test
4.5
20
Image to Video (Kling 2.5 Pro)
1 test
4.5
21
Image to Video (Kling 2.6 Pro)
1 test
4.0
22
Reference to Video (Kling O3 Pro)
1 test
4.0
23
Seedance Pro 1.5
1 test
4.0
24
Sora 2
1 test
3.0