Compare AI image models
Pick up to three models and compare them on the same design-production prompts: overall score, per-criterion quality, cost and latency, side by side. Your selection is saved in the URL, so any comparison is shareable.
Select two or three models to compare.
pilot-0 dataset.
Featured head-to-heads
Hand-picked matchups where the trade-off is worth spelling out: the ones product teams actually decide between.
FLUX.2 vs Stable Diffusion 1.5
Still running Stable Diffusion in your stack? FLUX.2 scores 3.05 to SD 1.5’s 1.46 across the six criteria that describe the file, and renders clean type where SD garbles it, at $0.012 per image via API. The self-hosting control question against a measured, two-times quality gap.
Nano Banana 2 vs Seedream 4.5
Two mid-cost routing targets that have drifted apart. Seedream 4.5 now tops 13 of the 37 prompts and reaches a perfect 5.00 on resolution, while Nano Banana 2 tops none and returns a quarter of the pixels. The value decision for volume generation.
FLUX.2 vs Nano Banana 2
The two default modern routes, measured against a field that has since grown to 23 models. FLUX.2 is the faster and cheaper of the pair at $0.012 and a 2.0 second median; Nano Banana 2 answers with better adherence and composition at roughly seven times the price. Neither wins a prompt outright any more.
FLUX.2 vs GPT Image 1.5
The premium-vs-value routing call. FLUX.2 costs roughly twelve times less and is the second-fastest model measured; GPT Image 1.5 leads the suite on file quality at 4.16 and is the only model that emits a usable alpha channel. Many pipelines need both, routed per job.
GPT Image 1.5 vs Nano Banana 2
OpenAI vs Google at the top of the file-quality table. GPT Image 1.5 owns true cut-outs, the only usable alpha channel in the field, and sits at the 5.00 text ceiling; Nano Banana 2 is the cheaper generalist that no longer leads on brand color. Which flagship earns the expensive lane in your router.
Nano Banana 2 vs Nano Banana Pro
Is Pro worth roughly twice the price in production? Nano Banana Pro costs $0.150 per image, runs slower, and scores lower on our blended ranking than the mainline model (3.13 vs 3.34). The unit-economics check before you default an entire feature to the Pro tier.
FLUX.2 vs Ideogram 3.0
The text-rendering reputation, tested. Ideogram built its name on typography, yet it wins no prompt outright and its text accuracy of 4.71 is twentieth of 23, below FLUX.2. Verify reputations against current measurements before you route text-critical work.
Ideogram 3.0 vs Recraft V3
The two models pitched at design teams, head to head. Both land below their reputation on this suite, finishing nineteenth and twentieth on the blended ranking, and Recraft V3 is twenty-second of 23 on brand color, adherence and composition alike, exactly the jobs its positioning claims. Recraft V4.1 now supersedes it on both score and price.
Nano Banana 2 vs Stable Diffusion 1.5
A current flagship against the legacy baseline, 3.40 against 1.46 across the six file criteria. The concrete measure of what migrating an SD-era integration to a modern hosted model buys you in output quality.
FLUX.2 [dev] Turbo vs Muse Image
The two cheapest models anyone would ship, and the file quality between them is not close. Turbo is $0.008 and returns in two seconds. Muse Image is $0.010, takes eighteen seconds, and buys 3.60 against 2.98 across the six file criteria, which puts it third on the blended ranking against Turbo in ninth. Latency against quality at the bottom of the price list.
GPT Image 1.5 vs GPT Image 2
The successor that went backwards for print. GPT Image 2 takes the best brand color in the field at 4.11, but it drops the alpha channel its predecessor had, cuts the output from 1.05 to 0.79 megapixels, costs more at $0.155, and takes 110 seconds against 34. Upgrading the model id is not an upgrade for every pipeline.
Recraft V3 vs Recraft V4.1
The design-tool upgrade, measured. Recraft V4.1 lifts every quality criterion its predecessor sat twenty-second on, 4.70 adherence against 3.31 and a clean 5.00 on text against 4.00, keeps the same partial alpha channel, and costs less at $0.035. That lifts it from twentieth to second on the blended ranking, and the only thing it gives up is three seconds of latency.
Seedream 4.5 vs Seedream 5.0 Pro
Is the new flagship worth three and a half times the price? Seedream 5.0 Pro builds the better file at 3.88 against 3.57 and leads on brand color and composition, but the two are level on resolution at 4.19 megapixels and Seedream 4.5 answers in twelve seconds against 65. Pro for hero work, 4.5 for everything that ships in volume.
FLUX.2 [dev] Turbo vs Seedream 4.5
The high-throughput tier, head to head. FLUX.2 Turbo is the cheaper and faster of the two at $0.008 and a 2.0 second median, while Seedream 4.5 buys five times the pixels and tops 13 of the 37 prompts on file quality for $0.040. For in-app generation and automation pipelines where unit cost rules.
New model? New benchmarks.
We re-run the identical 37-prompt suite on every model release. Get the scores, and what changed in the rankings, in your inbox.
You're on the list.
We'll email you when new benchmark results ship.