CL-SN55 941 —
GPT-61S 885 —
GPT-56T 861 —
GPT-6S 825 —
CL-OP55X 819 —
MUSE-SPK 793 —
QWEN-38X 779 —
GPT-6A 778 —
CL-OP5H 712 —
CL-OP5X 710 —
KIMI-K3X 700 —
CL-OP46H 695 —
CL-FAB5H 693 —
CL-OP55H 693 —
CL-OP47H 684 —
GEM-37FH 649 —
GEM-38FH 634 —
GPT-56S 570 -0.5%
CL-OP47 537 -0.4%
INKL 531 —
GEM-3P 494 —
CL-OP46 489 -0.2%
CL-OP48 482 -0.2%
GEM-31P 457 -0.2%
GPT-6L 436 —
CL-SN55 941 —
GPT-61S 885 —
GPT-56T 861 —
GPT-6S 825 —
CL-OP55X 819 —
MUSE-SPK 793 —
QWEN-38X 779 —
GPT-6A 778 —
CL-OP5H 712 —
CL-OP5X 710 —
KIMI-K3X 700 —
CL-OP46H 695 —
CL-FAB5H 693 —
CL-OP55H 693 —
CL-OP47H 684 —
GEM-37FH 649 —
GEM-38FH 634 —
GPT-56S 570 -0.5%
CL-OP47 537 -0.4%
INKL 531 —
GEM-3P 494 —
CL-OP46 489 -0.2%
CL-OP48 482 -0.2%
GEM-31P 457 -0.2%
GPT-6L 436 —
← Back to feed

Microsoft MAI-Image-2.6 Debuts at #2 on Arena Image Edit, Ahead of Google, Meta, and xAI

Microsoft AI Image 2.6 entered the Arena Image Edit leaderboard at number two on August 18, 2026 — ahead of image models from Google, Meta, and xAI at debut. Arena’s Image Edit category tests models on instruction-following image modification tasks, not text-to-image generation; it is a more operationally relevant benchmark for the enterprise use cases Microsoft is targeting.

The result is a step forward from the previous generation. MAI-Image-2.5 debuted at #3 in May 2026 on the Text-to-Image leaderboard. The new version climbs one position and shifts category, landing above a more competitive field in a leaderboard that now includes xAI’s Imagine Image 2.0, which entered at #2 in August.

Where It Sits

Arena Image Edit ranks models on human preference votes from side-by-side comparisons. A #2 debut means MAI-Image-2.6 won more head-to-head votes than Google’s image editing offering, Meta’s Imagine variant, and xAI’s Aurora in the same timeframe. Number one is not disclosed in the changelog entry.

For context on the competitive landscape: xAI’s Imagine Image 2.0 had entered at #2 in Arena’s Image Edit leaderboard as recently as August 9, trailing GPT-Image-2 by 24 Elo points in editing. Microsoft taking the same position with 2.6 indicates a tight cluster at the top.

Microsoft’s Image Model Trajectory

Microsoft’s MAI (Microsoft AI) image series has progressed steadily: three in-house models launched in April 2026 broke the lab’s dependence on OpenAI’s image generation pipeline. The 2.5 release in May, now the 2.6 release in August, represents a four-month development cadence.

Image editing — as opposed to generation — is where enterprise value concentrates: product photography adjustments, document image correction, and multimodal agent tasks that modify existing assets rather than synthesising new ones. Arena’s Image Edit leaderboard is the most direct signal of how models perform on that workload.

The full Elo score for MAI-Image-2.6 has not been published separately from the Arena changelog entry.