The best AI image generators, ranked.
82 text-to-image models scored from 2.9M blind, side-by-side human votes on Arena. Refreshed every week — so you always know which generators an AI detector has to keep up with.
- 82
- Models ranked
- 2.9M
- Human votes
- Oct 7, 2026
- Data as of
- 8
- Categories
- 1430±10
- 1414±10
- 1382±6
Photorealistic & Cinematic Imagery
Prompts asking for photographic or film-like realism — the category that matters most for detection.
2,910,195 votes · 82 models · as of Oct 7, 2026
Photorealistic & Cinematic Imagery rankings
| Rank | Model | Score | ||
|---|---|---|---|---|
1– | 1 – 2 | gpt-image-2.5-sunburstPreliminary OpenAI·Proprietary | 1430±10 | 6,435 |
2– | 1 – 2 | gpt-image-2.5-flarePreliminary OpenAI·Proprietary | 1414±10 | 5,913 |
3– | 3 | OpenAI·Proprietary | 1382±6 | 35,454 |
4NEW | 4 – 6 | gemini-nano-banana-2.1Preliminary Google·Proprietary | 1339±14 | 2,055 |
5▼1 | 4 – 6 | Microsoft AI·Proprietary | 1330±8 | 9,177 |
6NEW | 4 – 8 | grok-imagine-image-2.0 (canvas)Preliminary SpaceXAI·Proprietary | 1323±17 | 1,272 |
7▼1 | 6 – 8 | Reve·Proprietary | 1304±11 | 3,318 |
8▼3 | 6 – 8 | SpaceXAI·Proprietary | 1302±9 | 5,461 |
9▼2 | 9 – 11 | Meta·Proprietary | 1284±7 | 15,918 |
10▼2 | 9 – 11 | Bytedance·Proprietary | 1279±6 | 36,403 |
11▼2 | 9 – 13 | Reve·Proprietary | 1279±8 | 6,549 |
12▼2 | 11 – 15 | Google·Proprietary | 1266±6 | 23,958 |
13▼2 | 11 – 17 | Alibaba·Proprietary | 1262±10 | 4,392 |
14▼2 | 12 – 16 | Google·Proprietary | 1261±4 | 74,550 |
15▼2 | 12 – 18 | Google·Proprietary | 1260±8 | 7,439 |
16▼2 | 13 – 18 | Microsoft AI·Proprietary | 1252±6 | 23,678 |
17▼2 | 14 – 18 | OpenAI·Proprietary | 1248±4 | 67,969 |
18▼2 | 15 – 18 | Google·Proprietary | 1245±6 | 39,144 |
19▼2 | 19 – 20 | qwen-image-2.1Preliminary Alibaba·qwen-research | 1226±10 | 4,413 |
20▼2 | 19 – 22 | Alibaba·Proprietary | 1209±9 | 5,320 |
21▼2 | 20 – 22 | Ideogram·Ideogram open model | 1204±6 | 23,854 |
22▼2 | 20 – 22 | Microsoft AI·Proprietary | 1204±6 | 20,923 |
23▼2 | 23 – 26 | Luma AI·Proprietary | 1189±9 | 5,923 |
24▼2 | 23 – 26 | Luma AI·Proprietary | 1184±6 | 18,512 |
25▼2 | 23 – 31 | Recraft·Proprietary | 1180±18 | 1,014 |
Showing 25 of 82 models · data as of Oct 7, 2026
How settled is the top 20?
Arena scores come with a 95% confidence interval. Where intervals overlap, the order between two models is not statistically settled — that is exactly what the “rank spread” column in the table expresses.
Arena score with 95% confidence interval
Top 20 models · hover a rank for details
Average win rate against every opponent
Ties excluded · as published by Arena · a model can out-rank another yet win less often, because it has faced stronger opponents
- #178.2%gpt-image-2.5-sunburst
- #277.5%gpt-image-2.5-flare
- #365.8%gpt-image-2 (medium)
- #456.5%gemini-nano-banana-2.1
- #556.0%mai-image-2.6
- #753.7%reve-2.1
- #1153.5%reve-2.0
- #851.4%grok-imagine-image-2.0 (20260801)
- #649.2%grok-imagine-image-2.0 (canvas)
- #1547.3%gemini-3.1-flash-lite-image (nano-banana-2-lite)
- #946.9%muse-image
- #1846.3%gemini-3-pro-image-preview (nano-banana-pro)
- #1046.2%seedream-5.0-pro
- #1444.0%gemini-3-pro-image-2k (nano-banana-pro)
- #1243.8%gemini-3.1-flash-image (nano-banana-2) [web-search]
- #1340.3%qwen-image-3.0-pro
- #1738.8%gpt-image-1.5-high-fidelity
- #2035.9%qwen-image-2.0-pro-2026-06-22
- #1635.2%mai-image-2.5
- #1934.4%qwen-image-2.1
Every model on this board is a model our detector has to catch.
Sightova builds AI image detection. The generators people prefer are the generators that end up in fraudulent claims, fake profiles and manipulated news — so we watch the same leaderboard the labs do, and we make it public. Right now that means keeping pace with gpt-image-2.5-sunburst from OpenAI.
Photorealism is the threat
Switch to the Photorealistic or Portraits category: those rankings show which models produce the images most likely to pass as real photographs.
Metadata-independent
Our detection reads pixels, not watermarks or EXIF, so a new generator climbing this board is a training signal, not a blind spot.
Human-judged, not benchmarked
Arena ranks by anonymous side-by-side votes from real people. It measures what people actually prefer — and therefore what gets used.
Kept weekly
Snapshots are archived every week. Movement badges show who climbed, fell or entered since the last release.
How the scores are made
On Arena, a visitor types a prompt, two anonymous models each generate an image, and the visitor votes for the better one. Millions of those pairwise votes are fitted with a Bradley–Terry style model to produce each model's Arena score.
The ± figure is a 95% confidence interval; the rank spread is the range of ranks a model could hold once that uncertainty is taken into account. Preliminary marks models added in the last release, which still have comparatively few votes.
Sightova republishes the numbers as published — no re-weighting, no editorial adjustments. Scores are rounded to whole numbers for display.
Rankings, scores, confidence intervals and vote counts are data from Text-to-Image Arena · Photorealistic & Cinematic Imagery by Arena Intelligence, data as of Oct 7, 2026. © Arena Intelligence. Sightova is not affiliated with, sponsored by or endorsed by Arena; we mirror the board weekly with attribution. Model and lab names are trademarks of their respective owners.
View the live board on arena.ai