GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

Reve 2.1 Debuts at #2 on Text-to-Image Arena With Elo 1302 — Above Meta, Microsoft, and Google

Reve 2.1 entered Arena’s Text-to-Image leaderboard on July 9 and posted 1302 ELO in its first 2,432 battles — enough for second place. The score puts it 83 points behind GPT-Image-2’s 1385 and ahead of every major lab image model on the board.

Current Text-to-Image Rankings

RankModelVendorELOVotes
1GPT-Image-2 (medium)OpenAI138560,382
2Reve 2.1Reve13022,432
3Muse ImageMeta12808,384
4Reve 2.0Reve127113,675
5Gemini 3.1 Flash ImageGoogle126118,502
6MAI-Image-2.5Microsoft125734,416
7Gemini 3.1 Flash Lite ImageGoogle12509,392
8Gemini 3 Pro Image 2KGoogle1245129,467
11SeedDream 5.0 ProByteDance12312,865

Reve 2.1 leads Meta Muse Image by 22 points, Microsoft MAI-Image-2.5 by 45 points, and all Google Gemini image variants by 41 to 57 points. The only model above it is GPT-Image-2.

Reve also holds the fourth position through its 2.0 model. Two of the top four spots on the leaderboard belong to Reve, a lab that is not among the established AI infrastructure players.

Vote Counts and Confidence

The critical caveat is votes. GPT-Image-2 has 60,382 battles; Reve 2.1 has 2,432. Google Gemini 3 Pro Image 2K has 129,467 — more than 50x the Reve 2.1 count. Arena’s ELO scores converge as vote counts rise, which means a 1302 on 2,432 votes carries more uncertainty than a 1245 on 129,000.

Reve 2.0, with 13,675 votes, is the more established data point: it holds 1271 ELO, which puts Reve’s baseline above the Microsoft and Google image models even on a high-vote sample. Reve 2.1’s 31-point improvement over 2.0 is consistent with a meaningful quality jump, and it starts from a baseline that already beat most of the field.

ByteDance SeedDream 5.0 Pro

ByteDance also entered the Text-to-Image leaderboard on July 11 with SeedDream 5.0 Pro, posting 1231 ELO across 2,865 early votes — rank 11. SeedDream is ByteDance’s image counterpart to SeedDance (video), which holds both top positions on Arena’s Video leaderboard. SeedDream 5.0 Pro enters just above Grok Imagine Image Quality (1229) and below Google Gemini 3 Pro Image Preview (1232).

The SeedDream entry completes ByteDance’s Arena footprint across image and video categories. Its 1231 ELO is a solid debut but sits 71 points below Reve 2.1 and 154 points below the text-to-image leader.

What Reve 2.1 Represents

The leaderboard structure is significant regardless of whether Reve 2.1’s position holds as vote counts rise. Microsoft, Google, and Meta have all deployed image generation models with significant compute and research backing. A challenger reaching second place on human preference votes — beating all three of them in head-to-head Arena battles — suggests that the image generation quality gap between frontier labs and well-executed challengers has narrowed.

For the prior cycle, the Text-to-Image arena was effectively a three-way competition between OpenAI, Google, and Microsoft. Reve 2.1’s debut changes that framing. It will take another 10,000-plus votes before the position is statistically stable, but the early signal is not noise.