---
type: "post"
id: "best-ai-for-single-image-editing-image-edit-9e542f"
slug: "best-ai-for-single-image-editing-image-edit-9e542f"
title: "Best AI for Single-Image Editing: Image Edit Leaderboard"
description: "LMArena’s Single Image Edit leaderboard, dated November 2, 2026, places Microsoft AI’s MAI-Image-2.7-Preview fourth among 63 models."
retrieval_nugget: "LMArena’s Single Image Edit leaderboard, dated November 2, 2026, places Microsoft AI’s MAI-Image-2.7-Preview fourth among 63 models. The ranking offers a large-scale preference signal for builders comparing image-editing options, but it measures Arena votes rather than performance on any team’s own tasks."
published_at: "2026-11-06"
updated_at: "2026-11-06"
record_date: "2026-11-06"
date_kind: "published_at"
topics: ["models"]
entities: ["LMArena","MAI-Image-2.7-Preview","GPT-Image-2.5 Sunburst","GPT-Image-2.5 Flare","GPT-Image-2.6 Sunburst (Medium)","Microsoft AI"]
related_patterns: ["agent-economics-moves-to-completed-work"]
novelty_label: "new_fact"
what_changed: "LMArena’s image-edit leaderboard now ranks GPT-Image-2.5 Sunburst first (1517) and MAI-Image-2.7-Preview fourth (1469), a new Arena-specific ranking beyond the prior release and Artificial Analysis results."
prior_refs: ["https://newruntime.com/posts/microsoft-s-mai-image-2-7-preview-ranks-ccadfd/"]
source_url: "https://lmarena.ai/leaderboard/image-edit"
source_urls: ["https://lmarena.ai/leaderboard/image-edit"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:best-ai-for-single-image-editing-image-edit-9e542f"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/best-ai-for-single-image-editing-image-edit-9e542f/","markdown":"https://newruntime.com/posts/best-ai-for-single-image-editing-image-edit-9e542f.md","json":"https://newruntime.com/posts/best-ai-for-single-image-editing-image-edit-9e542f.json"}
markdown_url: "https://newruntime.com/posts/best-ai-for-single-image-editing-image-edit-9e542f.md"
---

# Best AI for Single-Image Editing: Image Edit Leaderboard

## Retrieval answer

LMArena’s Single Image Edit leaderboard, dated November 2, 2026, places Microsoft AI’s MAI-Image-2.7-Preview fourth among 63 models. The ranking offers a large-scale preference signal for builders comparing image-editing options, but it measures Arena votes rather than performance on any team’s own tasks.

## Summary
LMArena’s Single Image Edit leaderboard, dated November 2, 2026, places Microsoft AI’s MAI-Image-2.7-Preview fourth among 63 models. The ranking offers a large-scale preference signal for builders comparing image-editing options, but it measures Arena votes rather than performance on any team’s own tasks.

## What happened
LMArena’s Image Edit Arena page presents overall rankings for single-image editing. Its displayed date is November 2, 2026. The page reports 31,959,549 votes and 63 models. In the visible table, GPT-Image-2.5 Sunburst leads with a score of 1517±5 and 114,468 votes. GPT-Image-2.5 Flare follows at 1482±5 with 111,635 votes, and GPT-Image-2.6 Sunburst (Medium) is third at 1476±10, with 3,833 votes; the latter is marked preliminary.

Microsoft AI’s MAI-Image-2.7-Preview is fourth, at 1469±7 and 12,728 votes. The table identifies it as proprietary. Below it, GPT-Image-2.6 Flare (Medium) is fifth at 1462±10 and 3,850 votes, also marked preliminary. GPT-Image-2 (Medium) is sixth at 1460±3 with 359,999 votes. The ranking also lists Grok Imagine Image 2.0 seventh at 1438±6 and Google’s Gemini Nano Banana 2.1 eighth at 1427±6.

Microsoft’s earlier MAI-Image-2.6 appears at ninth place with a score of 1426±4 and 89,003 votes. Other entries include Krea 3 Ultra at eleventh, Meta’s Muse Image at thirteenth, and models from Google, ByteDance, Alibaba, Tencent, and others. The displayed scores are accompanied by uncertainty ranges, and several entries are explicitly labeled preliminary. The page provides model links but, in the supplied material, no description of the vote-collection procedure beyond the vote totals or details on access, pricing, editing features, or evaluation prompts.

The distribution of vote counts is uneven even near the top of the table. The three leading entries range from 3,833 to 114,468 votes, while MAI-Image-2.7-Preview has 12,728. Farther down, some entries have hundreds of thousands or more than a million votes. LMArena displays those counts alongside each score, so a reader can distinguish the leaderboard-wide total from the evidence shown for an individual model.

The result is therefore a ranking snapshot rather than a new model launch. MAI-Image-2.7-Preview sits close to the top of this particular preference table, while the page’s broad vote count does not mean every model has an equally large or mature sample. For a team choosing a model, the table can help identify candidates to test, not substitute for task-specific evaluation.

## What the archive adds
This is a new leaderboard result within a continuing image-editing comparison thread. Our earlier [MAI-Image-2.7-Preview note](https://newruntime.com/posts/microsoft-s-mai-image-2-7-preview-ranks-ccadfd/) reported it at #2 on AA-Image-Editing v2.1; [Krea 3 Ultra’s leaderboard note](https://newruntime.com/posts/krea-3-ultra-debuts-at-4-on-arena-1eb79b/) covered another set of editing rankings. The current source adds its LMArena Single Image Edit position, not a restatement of those exact benchmark results.

## Trend signal
The result supports a limited version of the [agent economics moves to completed work](https://newruntime.com/patterns/agent-economics-moves-to-completed-work/) signal: model selection is increasingly grounded in comparative task outcomes rather than model labels alone. But this leaderboard measures human preference on image edits, not verified task completion or cost, so it does not establish that the top-ranked option is the most economical production choice.

## Why now
The leaderboard is a dated snapshot, and its value is that it captures current relative positions as of November 2, 2026. The supplied material gives no explanation for why the ranking was updated at this time or what changed since an earlier snapshot.

## Why it matters
For teams shipping image-editing features, a high-ranking model deserves a place on the shortlist, not an automatic production switch. The spread between the top three and MAI-Image-2.7-Preview is visible, but the scores alone do not tell a team whether its users will prefer the same outputs, whether edits preserve required details, or what latency and cost will look like in its own stack.

The most useful response is to translate the public ranking into a small, representative test set. Include the tasks your product actually handles, compare candidate models blind where practical, and judge both preference and correctness. The leaderboard gives a starting point for that work; it does not settle it. Teams should also record the number of votes and preliminary labels when comparing positions, since a close score supported by a smaller sample carries more uncertainty than a result with a much larger vote count.

## Limits and uncertainty
The supplied page does not specify how prompts and source images are selected, how votes are recruited or weighted, or how the score is calculated. It does not establish statistical significance between nearby rankings, nor does it show results broken down by editing task, language, user group, or image type. The vote totals vary widely across entries, and several scores are labeled preliminary. The table also says nothing about API access, availability, price, latency, reliability, licensing beyond the model labels, or performance in a production workflow. Treat it as one comparative signal, not a universal quality verdict.

## Reader takeaway
MAI-Image-2.7-Preview is fourth in LMArena’s November 2 single-image-edit ranking, behind three GPT-Image entries. Use that as a reason to test it against your actual editing workload—not as proof it will outperform alternatives for your users. The leaderboard is most helpful for candidate discovery; deployment decisions still need local quality, cost, and reliability checks.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Models](https://newruntime.com/topics/models.md): Explore the models topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)