기본 콘텐츠로 건너뛰기

MAI Image 2.7 Pricing: Cost of 1,000 Images vs GPT Image 2.6

Summary: Microsoft AI's MAI Image 2.7 and MAI Image 2.7 Flash launched on November 17, 2026, and Microsoft puts them at roughly $35 and $10 per 1,000 images. We turned the per-token rates into per-image costs, checked them against OpenRouter's own sample runs, and lined them up against GPT Image 2.6 at every quality setting. Short version: Flash is the cheapest way into this tier, MAI Image 2.7 costs about the same as GPT Image 2.6 at low-to-medium quality, and the big saving only shows up against GPT Image 2.6 at its highest settings.

What launched, and where it is sold

Microsoft's launch post introduces two models: MAI Image 2.7, the higher-quality tier for design-ready visuals and careful edits, and MAI Image 2.7 Flash, the faster, cheaper tier for production volume. Both generate images from text and edit existing images. According to Microsoft, both support transparent backgrounds, and OpenRouter's model pages say both accept up to five reference images in one edit and can either use a fixed aspect ratio or pick one for the composition.

The models are in public preview on Microsoft Foundry and, per the launch post, through partners including fal, Genspark, OpenRouter and Vercel. On OpenRouter, each model currently has a single provider, Azure, so there is no price competition between hosts yet.

Microsoft's headline claim is that MAI Image 2.7 took the #1 spot on Artificial Analysis image editing "at just one-fifth of the cost" of GPT Image 2.6 Sunburst at max quality. The footnote gives the detail: as of November 16, 2026, 1,215 Elo across 5,478 votes, 4 Elo ahead of GPT Image 2.6 Sunburst (max). That is a vendor-reported ranking with a narrow margin, so read it as "roughly tied at the top", not as a clear win. We did not run our own quality test.

The published rates

MAI Image 2.7 and 2.7 Flash (OpenRouter, Azure provider)

  • MAI Image 2.7: text input $5.00, image input $8.00, image output $34.00 per 1 million tokens.
  • MAI Image 2.7 Flash: text input $1.75, image input $2.50, image output $10.00 per 1 million tokens.
  • Microsoft's own summary: about $35 per 1,000 images for MAI Image 2.7 and about $10 per 1,000 images for Flash. Both figures carry a "~", so treat them as approximations.

For context, OpenRouter still lists the previous generation: MAI Image 2.6 at $38 per 1M output tokens and MAI Image 2.6 Flash at $19. So the new Flash nearly halves the output rate of the old Flash, and the full model is a little cheaper than 2.6 too.

One gotcha: when we checked Azure's public Foundry Models pricing page on November 18, its Microsoft table listed MAI-Image-2, 2-Efficient, 2.5, 2.5 Flash and 2.5 Pro, but no 2.7 rows yet. If you buy directly on Foundry, confirm the 2.7 rate in your own Azure pricing view or invoice before budgeting.

GPT Image 2.6 Sunburst and Flare (OpenAI)

OpenAI's pricing page lists the same rates for both GPT Image 2.6 models: image input $8.00, cached image input $2.00 and image output $30.00 per 1M tokens, plus text input $5.00 (cached $1.25). The page also shows a Batch option at -50%. Sunburst is pitched for editing precision, Flare for speed. OpenAI's image guide says the two models can use different token counts at the same quality setting, even though the price per token is the same.

Turning tokens into cost per image

Token prices are hard to compare because each vendor turns an image into a different number of tokens. Here is how we converted each one. Our working is shown so you can redo it with your own numbers.

MAI Image 2.7 family

Microsoft does not publish a token count per image size for MAI Image 2.7, so we used two independent checks:

  1. Microsoft's figure: ~$35 per 1,000 images is $0.035 per image. At $34 per 1M output tokens, that is about 1,000 output tokens per image ($0.035 / $0.000034 = about 1,029, and a short text prompt accounts for a few hundredths of a cent of that). For Flash, $10 per 1,000 images at $10 per 1M is again about 1,000 tokens per image. That figure is our inference, not a documented spec.
  2. OpenRouter's sample runs: OpenRouter shows the cost of each of its benchmark prompts. For MAI Image 2.7, all ten text-to-image prompts cost $0.035 each, and its playground example (16:9, 1360x768) cost $0.03527. For Flash, all ten cost $0.010 each, with the example at $0.01045. That matches Microsoft's numbers closely.

Edits cost more because the input images are billed too. In the same OpenRouter runs, MAI Image 2.7 edits cost $0.041 to $0.043, and the four-reference edit cost $0.066. Flash edits cost $0.012 to $0.013, and the four-reference edit cost $0.020. So budget roughly $41 to $66 per 1,000 edits on MAI Image 2.7 and $12 to $20 on Flash, depending on how many reference images you send.

GPT Image 2.6

OpenAI's image generation guide includes a cost calculator for GPT-Image-2.6 (Sunburst and Flare). For a 1024x1024 image it estimates these output tokens per quality setting, which we multiplied by $30 per 1M:

  • low: 823 tokens = $0.0247 per image = $24.69 per 1,000
  • medium: 1,449 tokens = $0.0435 = $43.47 per 1,000
  • high: 1,756 tokens = $0.0527 = $52.68 per 1,000
  • xhigh: 3,122 tokens = $0.0937 = $93.66 per 1,000
  • max: 7,024 tokens = $0.2107 = $210.72 per 1,000

Those figures cover image output only. The calculator excludes text and image input tokens and the extra 100 output tokens per streamed partial image. OpenRouter's sample runs back the numbers up: GPT Image 2.6 Sunburst text-to-image prompts on "auto" quality cost between $0.025 and $0.078 each (an average of about $0.042 across the ten), and its edits cost $0.033 to $0.084.

Don't forget the prompt

Text input is billed separately. If your prompts average 100 tokens (our assumption for a typical one-paragraph prompt), that adds $0.50 per 1,000 images at $5 per 1M (MAI Image 2.7 and GPT Image 2.6) and about $0.18 at $1.75 per 1M (Flash). Long, detailed prompts change this only a little; reference images change it much more.

What 1,000 images cost, side by side

Option Free way to try Published price Approx. cost per 1,000 images Notable features
MAI Image 2.7 Flash Powers Canvas in MAI Labs (a limited preview; no price is shown there) $1.75 text / $2.50 image in / $10 image out per 1M ~$10 text-to-image; $12-20 edits (OpenRouter samples) Up to 5 reference images, transparent backgrounds, public preview
MAI Image 2.7 Listed in MAI Labs; no free API tier found $5 text / $8 image in / $34 image out per 1M ~$35 text-to-image; $41-66 edits Precision tier for iterative edits; #1 on AA image editing per Microsoft (4 Elo margin)
GPT Image 2.6 (Sunburst / Flare) No free API tier found $5 text / $8 image in / $30 image out per 1M; Batch -50% $24.69 (low) to $210.72 (max) at 1024x1024 Five quality levels, complexity setting, SVG and animated GIF output, markup-guided edits
Bing Image Creator (consumer app) 15 fast creations a day, then standard speed, up to 200 prompts per 24 hours Free; Microsoft Rewards points can buy more fast creations Not billed per image FAQ lists MAI-Image-2.6-Flash and MAI-Image-2e, not 2.7, as of Nov 18

Where the money actually goes:

  • Flash vs GPT Image 2.6 low: about $10 vs $24.69 per 1,000. Flash costs less than half of OpenAI's cheapest quality setting, before any batch discount.
  • MAI Image 2.7 vs GPT Image 2.6 medium/high: about $35 vs $43.47 to $52.68. Cheaper, but not dramatically. With OpenAI's Batch -50%, GPT Image 2.6 high drops to about $26.34 per 1,000, which is below MAI Image 2.7's list rate. Batch suits jobs that can wait, not interactive tools.
  • MAI Image 2.7 vs GPT Image 2.6 max: about $35 vs $210.72. This is the comparison behind Microsoft's "one-fifth of the cost" line. At 1024x1024 our math comes out closer to one-sixth.

A worked example: a shop that needs 5,000 product shots a month, each generated once and edited once with a single reference image, would pay about 5,000 x ($0.010 + $0.012) = $110 on Flash, or 5,000 x ($0.035 + $0.041) = $380 on MAI Image 2.7, using the OpenRouter sample costs. The same 10,000 outputs on GPT Image 2.6 at medium would be about $435 in output tokens alone, before the input image tokens for the edits.

Where to try it without an API bill

MAI Labs and Canvas. Microsoft's launch post says Canvas is a new workspace in MAI Labs powered by MAI Image 2.7 Flash, where you can start from a prompt or sketch and then replace, move or restyle any selected part of the image. The Labs page describes itself as a "limited preview of new and experimental AI models" and does not list a price, so we are not calling it free with fixed limits. Check what it shows you before relying on it.

Bing Image Creator. Microsoft says the MAI Image family is the default behind Bing Image Creator, and the Bing FAQ spells out the free allowance: 15 fast creations a day that refill daily, after which you drop to standard speed, with up to 200 prompts per 24-hour period (each "Edit image" click counts). However, on November 18 the FAQ's model list still named MAI-Image-2.6-Flash and MAI-Image-2e, not 2.7. Bing is a free way to get a feel for the model family, not a guaranteed preview of 2.7 itself.

Editor's note

Who this is for: teams generating images in volume (product shots, thumbnails, marketing variants) who want a cheaper tier than GPT Image 2.6 at high or max quality, and developers already on OpenRouter or Foundry who can switch with a model ID change.

Who it's not for: anyone who needs SVG or animated GIF output (OpenAI lists those for GPT Image 2.6), anyone who needs predictable regional pricing on Azure today (2.7 is not on the public Foundry price table yet), and anyone who needs a fixed free API quota.

Gotchas: "~$35 per 1,000" is for plain generation. Edits with several reference images cost up to about twice as much per output. Both MAI models are in public preview, so rates and limits can change.

Common errors and fixes:

  • Costs higher than the ~$10 or ~$35 you planned: check how many reference images each request sends. The four-reference sample cost $0.066 on MAI Image 2.7 vs $0.035 for text-only.
  • Your GPT Image 2.6 estimate doesn't match the bill: the calculator covers output tokens only, and "auto" quality changes per image. Set quality explicitly and read the token counts in the response's usage field.
  • Expecting 2.7 in Bing Image Creator: as of the FAQ we read, Bing lists 2.6 Flash and 2e. Use MAI Labs or the API if you need 2.7 specifically.

FAQ

How much does MAI Image 2.7 cost per image?
About $0.035 for text-to-image (Microsoft says ~$35 per 1,000; OpenRouter's samples were $0.035 each). Edits cost about $0.041 to $0.066, depending on the number of reference images. Flash is about $0.010 per image, or $0.012 to $0.020 for edits.

Is MAI Image 2.7 cheaper than GPT Image 2.6?
It depends on the quality setting you compare it with. Against GPT Image 2.6 at max (about $0.21 per 1024x1024 image), yes, by roughly six times. Against GPT Image 2.6 at medium or high ($0.043 to $0.053), it is only modestly cheaper, and GPT Image 2.6 with Batch -50% can undercut it. Flash is cheaper than every GPT Image 2.6 setting.

Can I use MAI Image 2.7 for free?
There is no free API tier that we found. Canvas in MAI Labs runs on 2.7 Flash as a limited preview without a listed price. Bing Image Creator is free (15 fast creations a day, then standard speed), but its FAQ currently lists MAI-Image-2.6-Flash and MAI-Image-2e rather than 2.7.

Related on lightmind24: OpenRouter Pricing: Free Models, 5.5% Fee and BYOK · Replicate vs fal.ai vs Fireworks: API Costs · Gemini Omni 1.2 Flash Pricing: What a 15-Second Clip Costs

If you generate images by the thousand, start on MAI Image 2.7 Flash and switch to the full model or GPT Image 2.6 only for the shots that need it.

Images: screenshots of Microsoft AI's launch post, OpenRouter's MAI Image 2.7 and 2.7 Flash provider pricing, OpenAI's API pricing page and the Bing Image Creator FAQ, plus our own cost table; see the official screenshots at the end of this post.

lightmind24 is an independent review site, not affiliated with Microsoft.

Last verified: 2026-11-18
Sources:
Microsoft AI: Introducing MAI Image 2.7 (Nov 17, 2026)
OpenRouter: MAI-Image-2.7
OpenRouter: MAI-Image-2.7 Flash
OpenRouter: GPT Image 2.6 Sunburst
OpenAI API pricing
OpenAI image generation guide (cost calculator)
Bing Image Creator FAQ
Azure Foundry Models pricing

Uploaded Image Uploaded Image Uploaded Image Uploaded Image

댓글

이 블로그의 인기 게시물

브루(Vrew) 요금제 정리: 무료 200크레딧으로 어디까지 쓸 수 있나

요약: 영상 자막 편집기 브루(Vrew)는 2026년 4월 22일부터 기능별 사용량 대신 '통합 크레딧' 하나로 요금을 매겨요. 무료 플랜은 매달 200크레딧이 들어오는데, 음성 분석(자동 자막)만 쓰면 최대 200분 분량이라 짧은 영상 몇 편은 충분하지만, 영상 초반의 Vrew 마크를 지우려면 유료 플랜이 필요해요. 이 글은 브루 공식 가격 페이지와 고객센터 공지를 2026년 10월 23일에 다시 확인해, 크레딧 계산법과 플랜 고르는 기준을 정리했어요. 목차 브루 요금 구조: 통합 크레딧이란 1크레딧으로 할 수 있는 일 무료 플랜으로 되는 것과 안 되는 것 유료 플랜 4종 비교표 월 구독·연 구독·기간 이용권, 무엇이 이득일까 상황별 추천: 학생·직장인·유튜버 크레딧이 모자랄 때: 사용 기록 지우개 업그레이드·환불 전에 알아둘 규칙 운영자 메모, 자주 묻는 질문 1. 브루 요금 구조: 통합 크레딧이란 예전 브루는 플랜마다 음성 분석 몇 분, AI 목소리 몇 자, 번역 몇 자처럼 기능별 한도를 따로 줬어요. 그런데 보이저엑스(브루 운영사)가 실제 이용 데이터를 보니 한두 가지 기능만 집중해서 쓰는 사람이 훨씬 많았다고 해요. 그래서 2026년 4월 22일 업데이트부터는 플랜마다 정해진 크레딧을 받고, 어떤 기능에 쓸지는 내가 정하는 방식으로 바뀌었어요. 플랜별 매달 지급량은 Free 200 · Light 2,000 · Standard 10,000 · Business 12,000크레딧 이에요. 공식 공지에 따르면 무료 사용자는 매달 1일에, 유료 사용자는 매월 결제일에 충전되고, 남은 크레딧은 다음 달로 넘어가지 않아요. 바뀐 뒤 무엇이 좋아졌나 공지에 나온 예시가 이해하기 쉬워요. Standard 플랜에서 AI 목소리만 쓰면 예전에는 최대 50만 자였지만, 지금은 1만 크레딧을 모두 AI 목소리에 몰아 쓸 수 있어 최대 100만 자가 돼요. 반대로 여러 기능을 골고루 쓰는 사람은 크레딧이 기능끼리 ...

에이닷 전화 vs 익시오: 통화 녹음·AI 요약, 내 통신사·폰에 맞는 앱 고르기

요약: 통화 녹음과 AI 요약을 쓰고 싶다면 먼저 내 통신사와 휴대폰 종류부터 확인해야 해요. LG유플러스 고객은 무료 앱 익시오, SKT 고객은 에이닷 전화가 가장 기능이 많고, 다른 통신사의 아이폰 사용자는 iOS에 들어 있는 기본 통화 녹음·전사 기능이 현실적인 선택이에요. 이 글은 2026-10-30에 SKT·LG유플러스·Apple 공식 페이지와 도움말을 직접 확인해 세 가지를 비교했어요. 왜 '앱 기능'보다 '내 조건'을 먼저 봐야 할까요 통화 요약 앱 비교 글을 보면 기능 목록부터 나열하는 경우가 많아요. 그런데 실제로 써 보려는 사람에게 가장 먼저 걸리는 건 기능이 아니라 이용 조건 이에요. 익시오는 LG유플러스 고객 전용이고, 에이닷 전화는 누구나 설치할 수 있지만 통역콜·baro 같은 기능은 SKT 고객에게만 열려 있어요. 아이폰은 운영체제 구조상 앱이 통화를 마음대로 녹음하기 어려워서, 통신사 앱이 안 되면 iOS 기본 기능에 기대야 해요. 그래서 이 글은 "어떤 앱이 제일 좋은가"보다 "내 통신사·내 폰에서 무엇을 쓸 수 있나" 순서로 정리했어요. 기능 차이는 그다음이에요. 한눈에 비교: 에이닷 전화 vs 익시오 vs 아이폰 기본 통화 녹음 구분 에이닷 전화 (SKT) 익시오 ixi-O (LG유플러스) 아이폰 기본 통화 녹음 (Apple) 누가 쓸 수 있나 통신사 구분 없이 앱 설치 가능. 단, 통역콜·baro·그룹통화 발신 등은 SKT 고객 전용 U+ 모바일 고객. 타 통신사, 만 14세 미만, 선불요금제 고객은 제한 (공식 사이트 FAQ상 알뜰폰 중 KB리브엠은 예외) 지원 국가·언어의 iPhone 사용자. 1명과 통화할 때만 녹음 가능 지원 OS·기기 Android 10 이상, iOS 16 이상 iOS 17 이상 iPhone 12 이상 / Android...

타입캐스트 vs 클로바더빙: 수익 나는 채널에 쓸 때 무료·유료 조건 비교

요약 타입캐스트와 클로바더빙은 둘 다 글만 입력하면 한국어 AI 목소리를 만들어 주는 서비스지만, 무료로 쓸 수 있는 범위가 꽤 다릅니다. 클로바더빙 무료는 매달 15,000자까지 쓸 수 있는 대신 수익이 생기는 채널에는 쓸 수 없고, 타입캐스트 무료는 처음에 주는 약 5분 분량의 다운로드로 끝나는 체험판에 가깝습니다. 광고 수익이 붙은 유튜브나 쇼핑몰 영상에 넣을 목소리라면 처음부터 유료 플랜 조건을 보고 고르는 편이 안전합니다. 왜 이 두 서비스를 같이 비교하나 쇼츠 내레이션, 수업 자료 더빙, 상품 소개 영상처럼 "내 목소리 대신 AI 목소리"가 필요할 때 한국 사용자들이 가장 많이 찾는 곳이 네오사피엔스의 타입캐스트 와 네이버의 클로바더빙 입니다. 기능 소개만 보면 비슷해 보이지만, 실제로 문제가 되는 건 음질보다 "이 목소리를 내 채널에 써도 되나" 입니다. 그래서 이 글은 기능 나열보다 무료 사용 범위, 출처 표기 의무, 유료 플랜 가격과 다운로드 한도를 중심으로 정리했습니다. 모든 수치는 2026년 10월 25일에 두 서비스의 공식 요금 페이지와 고객센터 도움말에서 직접 다시 확인한 내용입니다. 한눈에 보는 비교표 항목 타입캐스트 무료 타입캐스트 베이직 클로바더빙 Free 클로바더빙 Standard 월 요금 0원 9,900원 (연간 결제 시 10% 할인) 0원 19,900원 (네이버 클라우드 플랫폼에서 결제) 무료·포함 사용량 음성 생성·재생 무제한, 다운로드는 최초 3,000크레딧(약 5분) 1회 지급 월 30,000크레딧(약 60분) 월 15,000자, 다운로드 월 20회, 프로젝트 누적 5개 (매월 1일 갱신) 월 30,000자, 다운로드 월 40회, 프로젝트 누적 20개 쓸 수 있...