Text-to-Image Leaderboard
This page shows the top 30 models on the Arena AI leaderboard, snapshotted daily so you can see how recent models are doing at a glance. See the Arena AI site for the full leaderboard, filters and latest changes.
This leaderboard measures how well text-to-image models turn a text prompt into an image.
Data updated: 2026-10-16 13:02:49 UTC / 2026-10-16 21:02:49 CST (Beijing time)
Top 30
| Rank | Rank Spread | Model | Score | Votes |
|---|---|---|---|---|
| 1 | 1-1 | gpt-image-2.5-sunburst 1 | 1425 (±7) Preliminary | 17,752 |
| 2 | 2-2 | gpt-image-2.5-flare 1 | 1398 (±7) Preliminary | 16,889 |
| 3 | 3-3 | gpt-image-2 (medium) 1 | 1383 (±4) | 93,334 |
| 4 | 4-6 | grok-imagine-image-2.0 (canvas) 1 | 1334 (±12) Preliminary | 3,113 |
| 5 | 4-6 | mai-image-2.6 1 | 1332 (±5) | 25,091 |
| 6 | 4-6 | gemini-nano-banana-2.1 1 | 1328 (±9) Preliminary | 5,310 |
| 7 | 7-8 | reve-2.1 1 | 1302 (±8) | 8,249 |
| 8 | 7-8 | grok-imagine-image-2.0 (20260801) 1 | 1294 (±6) | 14,483 |
| 9 | 9-10 | muse-image 1 | 1274 (±5) | 42,262 |
| 10 | 9-11 | reve-2.0 1 | 1269 (±6) | 15,773 |
| 11 | 10-14 | gemini-3.1-flash-image (nano-banana-2) [web-search] 1 | 1262 (±4) | 60,515 |
| 12 | 11-15 | seedream-5.0-pro 1 | 1255 (±4) | 94,181 |
| 13 | 11-16 | qwen-image-3.0-pro 1 | 1255 (±6) | 11,440 |
| 14 | 11-16 | mai-image-2.5 1 | 1254 (±4) | 61,265 |
| 15 | 12-16 | gemini-3.1-flash-lite-image (nano-banana-2-lite) 1 | 1250 (±6) | 17,804 |
| 16 | 13-16 | gemini-3-pro-image-2k (nano-banana-pro) 1 | 1248 (±3) | 175,734 |
| 17 | 17-18 | gpt-image-1.5-high-fidelity 1 | 1237 (±3) | 169,927 |
| 18 | 17-19 | gemini-3-pro-image-preview (nano-banana-pro) 1 | 1234 (±5) | 86,464 |
| 19 | 18-19 | qwen-image-2.1 1 | 1223 (±7) Preliminary | 11,441 |
| 20 | 20-20 | ideogram-4.0-quality 1 | 1205 (±4) | 60,591 |
| 21 | 21-24 | qwen-image-2.0-pro-2026-06-22 1 | 1192 (±6) | 13,096 |
| 22 | 21-24 | uni-1.1-max 1 | 1189 (±6) | 14,072 |
| 23 | 21-25 | uni-1.1 1 | 1184 (±4) | 46,699 |
| 24 | 21-25 | mai-image-2 1 | 1184 (±5) | 51,475 |
| 25 | 25-27 | grok-imagine-image 1 | 1170 (±3) | 272,085 |
| 26 | 23-31 | recraft-v4.1-utility-pro 1 | 1169 (±11) | 2,617 |
| 27 | 25-31 | Cosmos3-Super-Text2Image (Agentic) 1 | 1165 (±7) | 10,805 |
| 28 | 26-31 | grok-imagine-image-pro 1 | 1162 (±4) | 99,455 |
| 29 | 26-31 | flux-2-max 1 | 1162 (±3) | 123,203 |
| 30 | 26-33 | flux-2-flex 1 | 1157 (±3) | 156,923 |
How to read this table
- Rank / Rank Spread: relative position estimated from text-to-image battles. The spread shows how far the rank moves within the confidence interval.
- Score: a relative score, useful for comparing the models listed at the same moment.
- Votes: sample size. Rankings move more when there are fewer votes.
Three things to check before choosing a model
- Whether the provider actually serves the model, and whether it is available in your region and for your account;
- Whether the API price, rate limits and context window fit your workload;
- Run a small test on three to five real tasks instead of relying on the overall ranking alone.
Data source
Data comes from the Arena AI Text-to-Image Leaderboard and is refreshed daily by GitHub Actions. For model pricing, licenses and capabilities, check the model provider's own documentation.