Klein 4B takes four references against eight on the other two, and offers one extra aspect ratio. **FLUX.2 Max and FLUX.2 Pro are identical on every setting the picker exposes** — same references, same aspect ratios, no variations, and no resolution field on either. Only the descriptions separate them.
Three models, and the interesting answer is that two of them can't be distinguished from the interface at all.
What differs?
Two rows, and both are on Klein.
| FLUX.2 Max | FLUX.2 Pro | FLUX.2 Klein 4B | |
|---|---|---|---|
| References | 8 | 8 | **4** |
| Aspect ratios | 8 | 8 | **9, including auto** |
| Resolution or quality | **No field** | **No field** | **No field** |
| Variations | None | None | None |
| Accepted types | Images | Images | Images |
| Description | "Maximum-detail rendering, reference-matched" | "High-detail FLUX rendering" | "Rapid-fire FLUX for quick exploration" |
Source: Hexcoded Creative Studio picker, Image → Create, September 2026.
Read the first two columns and they're the same column twice. Eight references, eight aspect ratios, no variations, no resolution or quality field.
Klein 4B differs on two axes. Half the references — four against eight. And one extra aspect ratio, because it's the only FLUX model offering auto.
So the "what do you give up on a cheaper tier" question has a clean answer for Klein and no answer at all for Pro.
FLUX.2 Max and FLUX.2 Pro are identical on every setting the picker shows. The descriptions are the only difference available to you.
Where this falls short. That's what the interface exposes, not what the models do. Max's description is "maximum-detail rendering, reference-matched" against Pro's "high-detail FLUX rendering," which implies a rendering difference. That isn't a setting, and we haven't measured it.
Why does no FLUX model have a resolution setting?
Unknown, and FLUX is the only family on the image roster where that's true of every member.
Most image models expose resolution tiers in K — 1K, 2K, 4K. The two OpenAI models use a quality toggle instead. All three FLUX models expose neither.
The panel goes references, prompt, aspect ratio, and then the generate button. There's no output-size decision to make.
Which raises a question the next section can't answer.
So why is Max priced at two resolutions?
This is the part we can't resolve, and it's worth stating rather than glossing.
Hexcoded's published cost panel lists FLUX.2 Max at 23 credits for 1K and 53 for 2K. Two resolution tiers, on a model with no resolution control in the picker.
So output size affects the price and isn't something you select. Whether it's determined by the aspect ratio, by the references, by something server-side, or whether the panel is simply out of date — none of that is visible from the interface.
FLUX.2 Pro and Klein 4B aren't on the panel at all, so there's no comparison to draw.
Where this falls short. Given that mismatch, the figures tell you roughly where FLUX.2 Max sits relative to other image models — cheaper than GPT Image 2, dearer than Seedream — and not what a specific generation will cost. The cost shown before you render is the authoritative one.
FLUX.2 Max is priced at two resolutions and has no resolution control. The picker can't tell you why.
What does Klein 4B actually give up?
Four references, and that's the whole answer.
Eight references on Max and Pro carries a character, a location, a style and a prop with room over. Four carries half of that — enough for a character and a look, not enough for a fully referenced composition.
For "quick exploration," which is what its description names, four is probably plenty. For reference-matched work, which is what Max's description names, it isn't.
What Klein gains is the auto aspect ratio, which neither of the other two offers. Minor, and it's the only thing on the roster Klein has that Max doesn't.
Where does FLUX sit on the image roster?
Mid-table on references, and unusual on everything else.
References: eight on Max and Pro puts them fifth of eleven — behind the two OpenAI models at sixteen, the Nano Banana family and Seedream 4.5 at fourteen, and Seedream 5.0 Pro at ten. Klein's four is second-lowest, ahead only of Recraft 4.1's one.
Aspect ratios: eight is mid-table. The Nano Banana family offers ten; GPT Image 1 Mini offers four.
Variations: none of the three batch. Only Recraft 4.1, Seedream 4.5 and GPT Image 2 do.
Resolution: the only family where no model exposes it.
So FLUX's distinguishing characteristic across the roster isn't a number — it's the absence of an output-size decision, which no other family shares.
So which should you use?
Two cases, and the third is genuinely undecidable from the settings.
Quick exploration
Klein 4B. Four references is enough for a character and a look, and its description names exploration. It's also the only FLUX model offering auto aspect ratio.
Reference-matched composition
Max or Pro — both take eight references, which is where these two separate from Klein. Which of the two is a question the picker can't answer.
Choosing between Max and Pro
Test them. Every setting is identical and the descriptions imply a rendering difference that isn't measurable from the interface. Generate the same prompt with the same references on both and compare.
All settings read from Hexcoded's Creative Studio picker in Image → Create mode in September 2026, all three models observed individually, including confirmation that none exposes a resolution or quality field. FLUX.2 Pro and Klein 4B do not appear on the published render costs panel, and FLUX.2 Max is listed there at resolution tiers the picker doesn't expose. Settings and costs change with model versions. The cost shown before you render is the authoritative figure.
- FLUX.2 Max and FLUX.2 Pro are identical on every setting the picker shows. Only the descriptions differ
- Klein 4B takes four references against eight. That's the one substantive difference in the family
- Klein also offers one extra aspect ratio — it's the only FLUX model with auto
- No FLUX model exposes a resolution or quality setting. The only family on the image roster where that's true
- FLUX.2 Max is priced at 1K and 2K on the cost panel despite having no resolution control
- Neither FLUX.2 Pro nor Klein 4B appears on the cost panel at all
- None of the three batch. Every generation produces one image
- Choosing between Max and Pro means testing them. The settings can't tell you
Nothing the picker exposes. Both take eight image references, both offer the same eight aspect ratios, neither has a variations field, and neither has a resolution or quality setting. Only the descriptions differ — "maximum-detail rendering, reference-matched" against "high-detail FLUX rendering."
Half the references — four against eight on the other two. That's the substantive difference. It gains one thing: the auto aspect ratio, which neither Max nor Pro offers.
The picker doesn't say. None of the three exposes a resolution or quality field, which makes FLUX the only family on the image roster where no model offers an output-size decision.
We can't tell you. The panel lists 23 credits for 1K and 53 for 2K, and the picker has no resolution control. So output size affects the price without being selectable, and nothing in the interface explains how it's determined. Use the cost shown before you render.
No. None of the three has a variations field. On the image roster only Recraft 4.1, Seedream 4.5 and GPT Image 2 batch — at sixteen, fifteen and ten respectively.
By testing. Every setting is identical, and the descriptions imply a rendering difference that isn't measurable from the interface. Generate the same prompt with the same references on both and compare the output.
Test the pair, pick the tier
Every setting visible per model before you generate, with the cost shown alongside. Eleven image models on the same credit balance as the video roster, so comparing two costs a decision rather than a subscription.
Open Creative StudioMore on model capability, access and rights in Models.