Quick answer
The current FLUX.2 pricing decision is not Dev vs. Pro vs. Schnell:
- Choose FLUX.2 [klein] for BFL’s fast, high-volume API path or supported open-weight deployment.
- Choose FLUX.2 [pro] for BFL’s production-grade hosted default.
- Choose FLUX.2 [flex] when typography and adjustable control matter.
- Choose FLUX.2 [max] when the highest-quality BFL tier or grounding search is required.
- Choose FLUX.2 [dev] for local development, then verify the exact license before commercial deployment.
BFL’s current FLUX.2 catalog does not list a FLUX.2 [schnell] tier. That older label should not be used as a current endpoint, price, or license shortcut.
This guide was source-checked on July 31, 2026. Prices, previews, licenses, and endpoint availability can change; use the live BFL pricing calculator and documentation before purchase or deployment.
Current official FLUX.2 price snapshot
Black Forest Labs prices hosted FLUX.2 requests by output resolution. Its documentation says one credit equals $0.01 USD and lists these starting prices:
| FLUX.2 option | Current BFL surface | Text-to-image starting price | Image-editing starting price | Best starting use |
|---|---|---|---|---|
| [klein] 4B | API and open weights | $0.014 | $0.014 | Fast or high-volume generation |
| [klein] 9B | API and open weights | $0.015 | $0.015 | A quality-and-speed balance |
| [pro] | Hosted API and Playground | $0.03 | $0.045 | Production workflows |
| [flex] | Hosted API and Playground | $0.05 | $0.05 | Typography and fine-grained control |
| [max] | Hosted API and Playground | $0.07 | $0.07 | Highest-quality tier and grounding search |
| [dev] | Local weights only | Local only | Local only | Development and custom local pipelines |
These are starting rates, not a quote for every image. BFL says FLUX.2 cost scales with output megapixels. Batch requests multiply the base cost by the number of images, and editing can have a different starting rate from text-to-image.
Use the official FLUX.2 pricing documentation for the current table and the BFL pricing calculator for the requested dimensions and inputs.
Pro vs. Klein: the practical choice
Choose Klein when speed or volume leads
BFL positions Klein for real-time and high-volume work. The 4B and 9B variants are separate models with separate licenses:
- Klein 4B uses Apache 2.0 according to the current BFL overview.
- Klein 9B uses the FLUX Non-Commercial License by default.
- Both have hosted API options, but local use still requires checking the model card and license that apply to the chosen weights.
Klein is a sensible first test for previews, interactive generation, large evaluation sets, or a workflow where many candidates will be rejected.
Choose Pro when a hosted production default leads
BFL describes Pro as its production-grade generation and editing option. The current API exposes both a preview and a fixed endpoint:
flux-2-pro-previewreceives BFL’s latest improvements.flux-2-prois a fixed snapshot for workflows that need reproducibility.
The preview endpoint may be the best place to evaluate current quality, but a pinned endpoint can be safer when regression tests, approvals, or compliance records depend on stable behavior.
Test accepted-image cost, not list price alone
A starting price cannot tell you which model is cheaper for your brief. Use the same prompt set, reference images, dimensions, and acceptance rubric, then record:
- requests submitted;
- billed megapixels and batches;
- failed or rejected outputs;
- edits, upscales, and retries;
- review time;
- images that pass the final requirement.
Use:
Accepted-image cost = total generation and editing spend ÷ accepted images
If Pro needs fewer retries for a demanding edit, it can beat a cheaper endpoint on the final unit. If Klein passes the same bar, its lower starting rate may matter at volume.
Where Flex, Max, and Dev fit
Flex
Use Flex when typography, small-detail preservation, or adjustable generation controls are central to the task. Compare it with Pro on the exact text, layout, and edit cases rather than assuming a general quality hierarchy.
Max
Use Max when BFL’s highest-quality tier or grounding search is required. Grounding can make current information available to a generation, but it does not replace source, rights, brand, or factual review of the finished image.
Dev
FLUX.2 Dev is not a cheaper hosted API tier. BFL’s current pricing documentation lists it as local only with no hosted API endpoint and a non-commercial default path.
Self-hosting also has costs that a per-image API table does not show:
- GPU purchase or rental and idle capacity;
- deployment, monitoring, scaling, and incident response;
- storage and delivery;
- moderation and human review;
- model updates, reproducibility, and rollback;
- commercial licensing where required.
BFL separately lists commercial open-weight licensing options. Do not infer commercial permission from downloadable weights, local execution, or a product-plan name.
What happened to Schnell?
Users may still search for “Schnell” because earlier FLUX families and third-party catalogs used that label. BFL’s current FLUX.2 overview instead uses Klein, Pro, Flex, Max, and Dev.
Before integrating, copy the exact endpoint from the current BFL API documentation. Do not convert a familiar family nickname into an API ID, price, performance promise, or license.
A safe budget worksheet
Estimate one representative workload rather than a hypothetical month:
| Input | Record |
|---|---|
| Task | Text-to-image or editing |
| Endpoint | Preview or fixed model ID |
| Output | Width, height, and billed megapixels |
| References | Count and size of input images |
| Batch | Images requested per call |
| Acceptance | Exact visual and policy checks |
| Iteration | Average attempts per accepted image |
| Operations | Review, moderation, storage, and delivery |
Then model low, expected, and high retry cases. Recalculate from actual API cost fields and accepted outputs after the pilot.
Related Flowith guides
Use the FLUX.2 model page for a concise model overview, the FLUX.2 licensing and API FAQ for deployment questions, or the AI image workspace to evaluate an image workflow.
Provider availability does not establish Flowith availability. Verify the current Flowith model selector and billing surface separately.
Frequently asked questions
How much does the FLUX.2 API cost?
Black Forest Labs listed FLUX.2 API text-to-image prices starting at $0.014 for Klein 4B, $0.015 for Klein 9B, $0.03 for Pro, $0.05 for Flex, and $0.07 for Max when checked July 31, 2026. FLUX.2 pricing scales with output resolution, so use the live BFL calculator for the exact request.
Is FLUX.2 Dev available through the hosted BFL API?
No. BFL’s current pricing documentation lists FLUX.2 Dev as local only, with no hosted API endpoint. Its default open-weight path is non-commercial; BFL separately offers commercial open-weight licenses, so verify the license for the exact deployment.
Is there a FLUX.2 Schnell API tier?
BFL’s current FLUX.2 model and pricing pages do not list a FLUX.2 Schnell tier. For fast, high-volume FLUX.2 generation, the current family uses the Klein name. Do not map an older Schnell label to a current endpoint or license.
Should I use FLUX.2 Pro or Klein?
Start with Klein when latency, volume, or local deployment matters. Start with Pro when a production-grade hosted default matters more. Test both on the same prompts and compare total cost per accepted image, not only the starting price.
Does Flowith access come with a BFL API purchase?
Do not assume so. BFL API credits, BFL open-weight licenses, and Flowith model availability are separate product surfaces. Check the current Flowith model selector and its billing terms independently.