As of 5 November 2026, no option is proven ahead. 2 options of 19 cannot be separated with the evidence in hand: GPT-Image-2.5 Sunburst and GPT-Image-2.6 Sunburst (Medium).
- Options
- 19
- Measured
- 19
- Shortlist
- 2
- Plotted
- 8
- Priced
- 8
The job
Generate pictures from a text prompt through a vendor's own API, and judge the model by how often people prefer its pictures in blind comparisons on Arena, against what 1,000 pictures cost.
Ranking
19 options measured, shortlisted rows shaded
| # | Option | Score 0 to 100 | Cost per 1,000 images | Interval | Price page |
|---|---|---|---|---|---|
| 1 | GPT-Image-2.5 SunburstopenaiShortlist | 85.4 | Not priced | ci9584.4 to 86.5 | |
| 2 | GPT-Image-2.6 Sunburst (Medium)openaiShortlist | 81.8 | Not priced | ci9579.2 to 84.4 | |
| 3 | GPT-Image-2.5 Flareopenai | 79.9 | Not priced | ci9578.9 to 80.9 | |
| 4 | GPT-Image-2.6 Flare (Medium)openai | 76.7 | Not priced | ci9574.2 to 79.2 | |
| 5 | GPT-Image-2 (Medium)openaiFrontier | 76.4 | $53.00 | ci9575.7 to 77.2 | |
| 6 | MAI-Image-2.6microsoft-ai | 66.1 | Not priced | ci9565.1 to 67.2 | |
| 7 | Gemini Nano Banana 2.1googleFrontier | 65.1 | $33.60 | ci9563.3 to 66.8 | |
| 8 | Grok Imagine Image 2.0xai | 63.5 | $40.00 | ci9562.0 to 64.9 | |
| 9 | Reve 2.1reve | 60.1 | Not priced | ci9558.6 to 61.6 | |
| 10 | Grok Imagine Image 2.0 (20260801)xai | 58.1 | Not priced | ci9557.1 to 59.2 | |
| 11 | Muse Imagemeta | 53.6 | Not priced | ci9552.7 to 54.5 | |
| 12 | Reve 2.0reve | 53.5 | Not priced | ci9552.3 to 54.8 | |
| 13 | Gemini 3.1 Flash Image (nano-banana-2, Web Search)google | 51.5 | $67.00 | ci9550.7 to 52.3 | |
| 14 | Qwen-Image-3.0-Proalibaba | 50.8 | Not priced | ci9549.5 to 52.1 | |
| 15 | MAI-Image-2.5microsoft-ai | 50.6 | Not priced | ci9549.8 to 51.4 | |
| 16 | Gemini 3.1 Flash-Lite Image (nano-banana-2-lite)google | 49.9 | $33.60 | ci9548.7 to 51.1 | |
| 17 | Grok Imagine ImagexaiFrontier | 33.9 | $20.00 | ci9533.4 to 34.4 | |
| 18 | Recraft V4recraft | 23.0 | $40.00 | ci9522.3 to 23.6 | |
| 19 | Recraft V4.1 FlashrecraftFrontier | 19.7 | $7.00 | ci9517.8 to 21.6 |
Order is by score, highest first. Intervals overlap, so neighbouring ranks do not differ with certainty. Cost is in USD per 1,000 images; bars start at zero. Interval: ci95, built from 95% confidence intervals given by the source.
Price against performance
Score against cost of 1,000 images, log scale
- Frontier: no other option is both cheaper and scores higher
- Other option
- Ring: on the shortlist, not separable with certainty
- Line: score interval, ci95 (built from 95% confidence intervals given by the source)
11 options not plotted: a coordinate is unknown
- GPT-Image-2.5 Flare: Cost unknown
- GPT-Image-2.5 Sunburst: Cost unknown
- GPT-Image-2.6 Flare (Medium): Cost unknown
- GPT-Image-2.6 Sunburst (Medium): Cost unknown
- Grok Imagine Image 2.0 (20260801): Cost unknown
- MAI-Image-2.5: Cost unknown
- MAI-Image-2.6: Cost unknown
- Muse Image: Cost unknown
- Qwen-Image-3.0-Pro: Cost unknown
- Reve 2.0: Cost unknown
- Reve 2.1: Cost unknown
Cost for this job
Date read and source for all 19 options
Cost of 1,000 images, in USD per 1,000 images, cheapest first within each group.
Workload and assumptions
Cost of 1,000 images of about one megapixel (1024x1024, called 1K by Google), at medium quality where the vendor prices by quality. Vendors that charge one flat price per image and state no size or quality (xAI, Recraft raster) are priced at that flat rate. Editing, upscaling, vector output and the web search of a model are not counted.
| Option | Cost | Price read | Source | Notes |
|---|---|---|---|---|
| Cost 8 | ||||
Recraft V4.1 Flashrecraft | $7.00 | 0 days old | ||
Grok Imagine Imagexai | $20.00 | 0 days old | ||
Gemini 3.1 Flash-Lite Image (nano-banana-2-lite)google | $33.60 | 0 days old | ||
Gemini Nano Banana 2.1google | $33.60 | 0 days old | ||
Grok Imagine Image 2.0xai | $40.00 | 0 days old | ||
Recraft V4recraft | $40.00 | 0 days old | ||
GPT-Image-2 (Medium)openai | $53.00 | 0 days old | ||
Gemini 3.1 Flash Image (nano-banana-2, Web Search)google | $67.00 | 0 days old | ||
| No cost available 11 | ||||
GPT-Image-2.5 Flareopenai | Unknown | Not read | No source | 1 note
|
GPT-Image-2.5 Sunburstopenai | Unknown | Not read | No source | 1 note
|
GPT-Image-2.6 Flare (Medium)openai | Unknown | Not read | No source | 1 note
|
GPT-Image-2.6 Sunburst (Medium)openai | Unknown | Not read | No source | 1 note
|
Grok Imagine Image 2.0 (20260801)xai | Unknown | Not read | No source | 1 note
|
MAI-Image-2.5microsoft-ai | Unknown | Not read | No source | 1 note
|
MAI-Image-2.6microsoft-ai | Unknown | Not read | No source | 1 note
|
Muse Imagemeta | Unknown | Not read | No source | 1 note
|
Qwen-Image-3.0-Proalibaba | Unknown | Not read | No source | 1 note
|
Reve 2.0reve | Unknown | Not read | No source | 1 note
|
Reve 2.1reve | Unknown | Not read | No source | 1 note
|
What evidence would settle it
The three conditions and what is missing
- GPT-Image-2.5 Sunburst: another independent origin or repeats (interval 2.1 points wide), metric Arena text-to-image rating (Elo-style, overall category).
- GPT-Image-2.6 Sunburst (Medium): another independent origin or repeats (interval 5.1 points wide), metric Arena text-to-image rating (Elo-style, overall category).
The three conditions for naming a best option
- The leader's lower bound must exceed every other option's upper bound by at least 3 points. Not met: the closest gap is 0.0 points (GPT-Image-2.5 Sunburst against GPT-Image-2.6 Sunburst (Medium)).
- Every critical metric must come from two independent origins, or from one auditable test of our own. Not met: 1 origin and no auditable own test for GPT-Image-2.5 Sunburst.
- The price must be verified within 30 days. Not met.
Sources and limits
5 sources, the method notes and this board as JSON
Scores on this board come from Arena Leaderboard Dataset (lmarena-ai/leaderboard-dataset). We rescale them to a 0 to 100 scale against fixed anchors (Arena text-to-image rating (Elo-style, overall category): 1000 counts as 0 and 1500 counts as 100), so a new option never moves another option's score. The licence and our changes are stated in the data credit at the foot of this page.
- Arena Leaderboard Dataset (lmarena-ai/leaderboard-dataset) free to use with credit; checked
- Gemini API pricing prices only, as facts with a link; checked
- OpenAI image generation guide prices only, as facts with a link; checked
- Recraft API pricing prices only, as facts with a link; checked
- xAI API models and pricing prices only, as facts with a link; checked
Notes on this board. Only one origin (Arena) measures quality, so no board entry can be named Best until a second independent origin or an own test exists. Arena's rating comes from blind votes by its own users on their own prompts; it is not a test of your job. Prices are read from each vendor's own API price page and are list prices for the stated workload; a model whose price could not be read on the vendor's page is listed with unknown cost and is not plotted. OpenAI's page prices the 2.5 and 2.6 image models by token and gives no per-image figure, so they have unknown cost. Selection: the 15 models with the highest Arena rank in the overall category, plus every model ranked 16 to 60 whose price was read on a vendor page and is at most $0.05 per image.
Scores are on a 0 to 100 scale against editorial anchors that are not yet validated. Prices are for the workload above and change without notice. Evidence older than 60 days is left out, and prices older than 30 days are flagged.
This board as JSON, with every input id, anchor and weight.