Model leaderboard

Every model runs the same Glif Standard Tests. Scores combine expert human reviewers, expert AI judges and votes from the wider community, and we rank models on quality, speed and cost.

Image · Image upscaling

Rated on 0 tests.

Score (50 is average)
Credits per run, log scale
Higher scores are better; further left is cheaper. Highlighted are the Glif defaults: GPT Image 2.6, Gemini 3.1 Flash.

  1. 1

    P-Image Upscale

    Pruna AI

    43 score out of 100, early score: expect it to move

    58 overall out of 100

  2. 2

    GPT Image 2.6Glif defaultNew

    OpenAI

    63 score out of 100, early score: expect it to move

    56 overall out of 100

  3. 3

    Gemini 3.1 FlashGlif default

    Google DeepMind

    39 score out of 100, early score: expect it to move

    55 overall out of 100

  4. 4

    Recraft Crisp Upscaler

    Recraft

    44 score out of 100, early score: expect it to move

    54 overall out of 100

  5. 5

    GPT Image 2.5

    OpenAI

    60 score out of 100, early score: expect it to move

    46 overall out of 100

  6. 6

    Clarity Upscaler

    51 score out of 100, early score: expect it to move

    31 overall out of 100

* Early score: not enough ratings yet, so expect it to move.