MAI-IMAGE-2.6 REACHES MICROSOFT FOUNDRY, AND A NEW MAI-IMAGE-2.6-FLASH JOINS IT (2026-09-04, vendor-primary): Microsoft published a **new, separately dated post** moving MAI-Image-2.6 out of the private preview this page recorded on 2026-08-10: '**Today, we're bringing it to developers in Microsoft Foundry and expanding the family with MAI-Image-2.6-Flash -- built to deliver that same level of quality for latency-sensitive, high-throughput production workloads.**' **This closes the 'no public rate card, private preview only' caveat that has sat on this page since August.** **NEW CAPABILITIES ON BOTH MODELS:** multi-image reference editing, web grounding, and dynamic aspect ratios. Microsoft frames the pair as a precision-versus-throughput choice: 'maximum precision with MAI-Image-2.6 and production speed with MAI-Image-2.6-Flash.' **MAI-IMAGE-2.6-FLASH PERFORMANCE, VENDOR-PUBLISHED:** '**2.8x faster than GPT-Image-2-Medium while delivering 72% greater efficiency**', at comparable quality. **THE RANK MOVED AGAIN, AND IT MOVED UP -- WHICH IS WHY THIS PAGE DATES ITS RANKS.** As of this post, MAI-Image-2.6 ranks **No. 2 on Arena for BOTH text-to-image and image editing** (as of 2026-09-04), and on **Artificial Analysis** it is **No. 2 for text-to-image and No. 1 for image editing**. **Track the editing rank across this page's own history, because it is a lesson in how fast these decay: 2.5 was No. 2 in June, had silently slipped to No. 5 by mid-August, 2.6 recovered it to No. 3 on 2026-08-18, and it is No. 2 as of 2026-09-04.** Treat every one of those as a dated snapshot, not a standing fact. Microsoft claims 'the best price-per-Elo performance in the world' -- a framing that is quality-per-dollar, not absolute quality, and should not be read as a claim to the top of either leaderboard. **A SOURCING NOTE FOR ANYONE RE-CHECKING THIS:** the URL that previously served the 2026-08-10 launch post now serves this 2026-09-04 post, with `datePublished` and `dateModified` both set to 2026-09-04 and no 'Updated as of' block. The August 18 editing-rank update recorded on this page is **no longer visible on the live page** -- it is preserved here rather than in Microsoft's copy.
2026-09-04•Microsoft AI (microsoft.ai/news/pushing-the-quality-cost-frontier-with-mai-image-2-6/, JSON-LD datePublished 2026-09-04) -- fetched 2026-09-05 via curl with browser UA; blog index enumerated via Firecrawl
STALENESS CATCH -- MAI-IMAGE-2.6 SHIPPED 2026-08-10 AND THIS PAGE HAD NEVER MENTIONED IT (recorded 2026-08-20, vendor-primary): a ten-day gap that three consecutive sweeps missed, because the vendor is Microsoft AI (microsoft.ai/news) rather than one of the newsrooms enumerated by default. **WHAT LAUNCHED:** MAI-Image-2.6, announced **2026-08-10**, '**ranked second on the Arena text-to-image leaderboard**', improving '**+79 Elo over MAI-Image-2.5 overall, with gains across every Arena text-to-image category**'. **Text rendering alone improves by +91 Elo.** Microsoft's own framing: the result 'firmly establishes MAI-Image ahead of leading models from Meta, Google and xAI'. **THE 2026-08-18 UPDATE TO THE SAME POST:** 2.6 also reached '**No. 3 on Arena's image editing leaderboard, ahead of Google's Nano Banana family and Meta's Muse Image**', a **+19 point gain over 2.5**, climbing **from No. 5 to No. 3**, with its biggest improvements in **Text Rendering (+43)** and **Product, Branding & Commercial Design (+38)**. **A DETAIL WORTH READING CAREFULLY RATHER THAN CELEBRATING:** this page previously described MAI-Image-2.5 as '**#2 image editing**' -- a June snapshot. The 8/18 update reveals 2.5 had **slipped to No. 5 on editing** before 2.6 pulled the family back to No. 3. So the leaderboard position in our old copy had silently decayed, and **2.6 on editing (No. 3) is a recovery to below where 2.5 originally sat, not a new high**. Text-to-image is the genuine improvement: No. 3 to No. 2. **AVAILABILITY IS NARROWER THAN 2.5's:** '**MAI-Image-2.6 is now available in MAI Playground and in Private Preview on Microsoft Foundry**' -- so unlike 2.5 it is **not** generally available on Foundry, and there is no mention of OpenRouter, PowerPoint, OneDrive or Bing Image Creator carrying it. **NO PRICING IS PUBLISHED ANYWHERE IN THE POST** -- treat 2.5's $5/$8/$47 as applying to 2.5 only. **BENCHMARK CAVEAT:** every number here is Arena Elo as reported by Microsoft in its own launch post; Arena is a third-party leaderboard, but the deltas and category breakdowns are Microsoft's presentation of it and no independent write-up is cited
2026-08-10•Microsoft AI (microsoft.ai/news/mai-image-2-6-launches-at-no-2-on-arena-ahead-of-google-meta-and-xai/, dated August 10, 2026, carrying an 'Updated as of August 18th, 2026' block, fetched 2026-08-20 via curl)
VERSION BUMP (2026-06-02, Microsoft Build): MAI-Image-2.5 launched as part of the 'seven new MAI models' wave. Vendor-published Arena results: #2 for image editing and #3 for text-to-image, with an overall Arena score 'surpassing Nano Banana Pro' and +75 points over MAI-Image-2 (text rendering +107, cartoon/anime/fantasy +90). New capabilities: advanced visual reasoning for scene understanding, fine-grained localized edit control, and face/identity consistency preservation across edits. Now live in PowerPoint and OneDrive (presentation visuals + photo editing), on Azure Foundry, MAI Playground, and via OpenRouter. Pricing: $5/1M text, $8/1M image-in, $47/1M image-out. The ultra-efficient MAI-Image-2.5-Flash ($1.75/$1.75/$19.50 per 1M) is NOW LIVE in Foundry alongside it (verified vendor-primary 2026-06-09).
2026-06-02•Microsoft AI (microsoft.ai/news/introducing-mai-image-2-5/, microsoft.ai/news/building-a-hillclimbing-machine-launching-seven-new-mai-models/)
Public preview on Azure Foundry -- availability is region-dependent. Global Standard deployment covers US + West Europe + Sweden Central + South India at launch. Other regions need to fall back to nearest available
2026-04•Microsoft Foundry catalog, Microsoft AI blog
Model card dated 2026-03-18 internally, publicly announced 2026-04-02 -- Microsoft has been running the model internally for several weeks before opening public preview, which explains the scale of Copilot/Bing integration at launch
2026-04•Microsoft model card PDF
Issues here are sourced from our editorial sweeps, not real-time telemetry. Newer issues may exist.