image model comparison
MAI Image 2.7 vs MiniMax H3-Image
MAI Image 2.7 and MiniMax H3-Image are both current image models, from Microsoft AI and MiniMax. On Arena Text-to-Image, MAI Image 2.7 scores 1387 Elo and MiniMax H3-Image 1198 Elo (Arena, measured November 17, 2026).
- Higher Arena Text-to-Image
- MAI Image 2.71387 Elo against 1198 Elo
- Newer release
- MAI Image 2.7Nov 17, 2026 against Nov 11, 2026
Data verified November 18, 2026 · Dataset 2026-11-18
Side by side
MAI Image 2.7 and MiniMax H3-Image, line by line
Prices in effect today, compared at the same tier wherever both list one. The better value in each row is highlighted.
| Fact | MAI Image 2.7 | MiniMax H3-Image |
|---|---|---|
| Price | ||
| Price per 1M input tokens | Not listed | $2.30 |
| Independent benchmarks | ||
| Arena Text-to-Image | 1387 Elo (#4 of 12) (better)Range 1375 to 1399 | 1198 Elo (#10 of 12)Range 1188 to 1209 |
| Arena Image Edit | 1475 Elo (#4 of 11) (better)Range 1464 to 1487 | 1358 Elo (#9 of 11)Range 1349 to 1367 |
| Specs | ||
| Provider | Microsoft AI | MiniMax |
| Release date | November 17, 2026 | November 11, 2026 |
| Access | API only | API and open weights |
| Context window | Not reported | Not reported |
| Max output | Not reported | Not reported |
| Input | text, image | text, image |
| Output | image | image |
Ranges under the scores are the publishers' own: published rating ranges from Arena. Where two ranges overlap, the gap is inside the measurement's own uncertainty, so the row says too close to call and neither score is highlighted or counted as a lead.
Scores:Arena · 2026-11-18
Charts
Against every current image model
Where MAI Image 2.7 and MiniMax H3-Image land among the current models of their task, on every independent benchmark, on price and on context.
Where each stands
On Arena Text-to-Image, MAI Image 2.7 ranks #4 and MiniMax H3-Image #10 of 12 current models. Each bar is the share of current models it beats: on score, price and context.
Independent benchmarks
Arena Text-to-Image
MAI Image 2.7:#4of 12
1,387 Elo · ahead of 73%
Published rating range 1,375 to 1,399
MiniMax H3-Image:#10of 12
1,198 Elo · ahead of 18%
Published rating range 1,188 to 1,209
Arena Image Edit
MAI Image 2.7:#4of 11
1,475 Elo · ahead of 70%
Published rating range 1,464 to 1,487
MiniMax H3-Image:#9of 11
1,358 Elo · ahead of 20%
Published rating range 1,349 to 1,367
Cost
What each costs
Example workloads at each provider's listed rates, then every other rate either one lists. The cheaper figure in each row is highlighted.
| Workload or rate | MAI Image 2.7 | MiniMax H3-Image |
|---|---|---|
| Other listed rates | ||
| 1M input tokens, image input | Not listed | $1.50 |
| 1M output tokens, image output | Not listed | $3.80 |
Your own bill will differ. These figures use list prices and example workloads. The AI Model Cost Audit reviews your real usage, shortlists models that score as well for each task for less, and estimates the savings.
Launch price $100, regular price $150
Verdict
When to choose each
MAI Image 2.7
Where it leads
- Higher Arena Text-to-Image score, 1387 Elo vs 1198 Elo
- Higher Arena Image Edit score, 1475 Elo vs 1358 Elo
Best for
- Edits that must keep the rest of a picture unchanged, such as swapping a product's color or material
- Product images, logos and icons that need transparent backgrounds
Watch out for
- Arena rates it below GPT Image 2.5 Sunburst on both boards (1428 and 1519) and below GPT Image 2.6 Sunburst for generation (1410), on 3,082 and 3,007 votes so far
- Microsoft has published only an approximate price, about $35 per 1,000 images, and no rate per image or resolution, so no price is recorded here
- A preview in Microsoft Foundry, under its preview terms
MiniMax H3-Image
Where it leads
- Open weights you can run yourself
Best for
- High-resolution product and marketing images where cost per image matters
- Teams that want to self-host or fine-tune an image model, where the license's territory rules allow it
Watch out for
- Arena rates it well below GPT Image 2.6 Sunburst: 1198 against 1410 on text-to-image and 1358 against 1481 on image editing (November 17, 2026), on 3,285 to 5,134 votes so far
- The weights' community license excludes the European Union, the United Kingdom, South Korea and the United States, and needs MiniMax's written authorization above $1 million in annual revenue or for any hosted service
- MiniMax publishes no quality benchmarks of its own
From benchmark to production
Building on MAI Image 2.7 or MiniMax H3-Image?
Public scores rarely hold on your own data. We test both models on your real tasks, then build the agent or pipeline around the one that wins.
AI Model Cost AuditLaunch price $100, regular price $150
FAQ
MAI Image 2.7 vs MiniMax H3-Image FAQ
Which scores higher on independent benchmarks, MAI Image 2.7 or MiniMax H3-Image?
MAI Image 2.7 leads on Arena Text-to-Image, 1387 Elo to 1198 Elo; MAI Image 2.7 leads on Arena Image Edit, 1475 Elo to 1358 Elo. Scores are independent evaluations, not either vendor's launch figures.
Can I run MAI Image 2.7 or MiniMax H3-Image myself?
MiniMax H3-Image is released as open weights you can run yourself; MAI Image 2.7 is available through its provider's API only.
AI Model Brief
The AI models worth knowing, once a week
What shipped, how it scores on independent benchmarks, what it costs per million tokens and what was deprecated, from the data behind our AI model benchmarks.
We use your address only to send the brief. See our privacy policy.
See this week’s briefFurther reading
More on MAI Image 2.7 and MiniMax H3-Image
Related comparisons
- GPT Image 2.6 Flare vs MAI Image 2.7
- Gemini 3.1 Flash Lite Image vs MiniMax H3-Image
- GPT Image 2.6 Sunburst vs MAI Image 2.7
- Gemini 3 Pro Image vs MiniMax H3-Image
- MAI Image 2.7 vs Nano Banana 2.1
- GPT Image 2.5 Flare vs MiniMax H3-Image
- GPT Image 2.5 Sunburst vs MiniMax H3-Image
- GPT Image 2.6 Flare vs MiniMax H3-Image
- GPT Image 2.6 Sunburst vs MiniMax H3-Image
Production evaluation
Choosing between MAI Image 2.7 and MiniMax H3-Image?
Public benchmarks narrow the shortlist. We run both on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.
- A head-to-head evaluation on your real workload
- Routing between models by cost and difficulty
