Model leaderboard

Ranked from 1007 blind head-to-head votes by 9 people on the Glif team, using the standard Glif tests. Each generation mode has its own ranking.

Ratings are Elo points on a scale centred on 1500, compared only within one mode; higher is better. Bars show 90% uncertainty ranges, so overlapping bars are close to a tie.

Image · Text-to-image

187 votes on 11 tests. A model needs 20 games to rank.

  1. 1

    Gemini 3.1 Flash Image Generator

    1702 Elo, 90% range 1612 to 1851

  2. 2

    Ideogram V4.5

    1670 Elo, 90% range 1567 to 1834

  3. 3

    Muse Image

    1610 Elo, 90% range 1497 to 1736

  4. 4

    FLUX.2 Pro Image Generator

    1607 Elo, 90% range 1519 to 1721

  5. 5

    Nano Banana Pro Image Generator

    1579 Elo, 90% range 1477 to 1722

  6. 6

    MAI Image 2.5 Generator

    1573 Elo, 90% range 1469 to 1694

  7. 7

    FLUX.2 Max Image Generator

    1556 Elo, 90% range 1445 to 1692

  8. 8

    Krea 2 Image Generator

    1506 Elo, 90% range 1410 to 1611

  9. 9

    Seedream V4.5

    1328 Elo, 90% range 1173 to 1454

  10. 10

    FLUX.2 Flex Image Generator

    1301 Elo, 90% range 1180 to 1398

  11. 11

    Seedream 5.0 Lite

    1294 Elo, 90% range 1121 to 1400

  12. 12

    Z-Image Generator

    1202 Elo, 90% range 1050 to 1308

  13. 13

    Grok Imagine Image

    1198 Elo, 90% range 1035 to 1299

Needs more ratings (2)

Quality and cost

  • Current default

Preference ↑ vs. 1500-rated reference

Cost → estimated credits per run, log scale

Predicted human preference within this mode, not an absolute quality grade. Cost is the midpoint of each estimated range, not a quote. Hover or focus for raw ratings.