FLUX Pro Ultra vs Recraft V3 for creator graphics

Match the proof to one real creator deliverable

A future authorized comparison would need one rights-cleared brief, exact IDs returned by Studio, a ratio proven compatible with both entries, disclosed settings, every unedited output, and a prewritten rubric. A possible brief structure is a fictional ceramic launch graphic with a named subject, framing, palette, and negative-space requirement. This is a proposed protocol only; no prompt was submitted and no controlled set exists.

A future rubric could inspect subject integrity, usable negative space, palette fit, observable artifacts, and crop resilience. It must record actual observations rather than pre-populate defects. Keep final words as live typography in the design tool. Recraft's first-party documentation discusses text positioning, but this research did not run exact-copy tests, and neither product positioning nor a vendor sample proves that a particular campaign line will render correctly.

Separate Studio choices from direct-provider controls

Black Forest Labs' direct documentation names `/flux-pro-1.1-ultra`, prices it at $0.06 per image, and exposes prompt, aspect ratio, Raw mode, seed, output format, image prompt, and related controls. Recraft's direct documentation prices V3 raster at $0.04 per image and V3 Vector at $0.08, while its style reference distinguishes raster styles from vector styles. Those pages describe provider products, not the exact input surface exposed by OfflineCreator Studio.

Published `@offlinecreator/mcp` 0.1.2 instead tells callers to obtain a Studio model ID from `list_models`. Its `generate` tool accepts `modelId`, `prompt`, optional `aspectRatio`, and optional `wait`; the published ratio enum is 16:9, 9:16, or 1:1. It exposes no Raw-mode, provider-style, vector-format, seed, or image-prompt field. Do not paste a provider endpoint name into `generate.modelId`, and do not promise that a direct-provider feature is available through Studio unless the current Studio model response and tool contract establish it.

Studio decision

Use the current list_models responseSelect the Studio model ID and published credit cost at run time.

Provider documentation

Capability context onlyIt does not prove that Studio exposes every direct-provider control.

Budget the comparison in Studio credits

At the current public Studio prices, one FLUX 1.1 Pro Ultra generation costs 8 credits and one Recraft V3 generation costs 5 credits. A minimal matched pair therefore costs 13 credits. Two candidates from each model cost 26 credits, and four from each cost 52 credits. These totals are arithmetic planning examples based on the cited catalog, not a recommendation to spend that amount and not evidence about output quality.

Direct-provider dollar prices are not Studio checkout prices. Black Forest Labs lists $0.06 per FLUX Ultra image, while Recraft lists $0.04 for V3 raster and $0.08 for V3 Vector. Studio publishes its own customer credit costs and labels wholesale figures as operating estimates. Refresh `list_models`, check the account balance, and approve a stop condition before generation. More samples reduce the risk of choosing one lucky output, but this page has no measured sample size that guarantees a stable model ranking.

One matched pair

28 Studio credits8 for FLUX Pro Ultra plus 5 for Recraft V3 at the cited catalog prices.

Two outputs per model

56 Studio creditsA small proof set, not a statistically validated benchmark.

Treat published samples as orientation, not the verdict

OfflineCreator's public model page labels its launch-model proof as original examples generated through the disclosed endpoints and shows cards for FLUX 1.1 Pro Ultra and Recraft V3. Its MCP page also presents Recraft V3 campaign and range examples. Black Forest Labs' Ultra documentation and Recraft's style documentation contain their own example galleries. These pages are useful for locating real, attributed samples, but they do not provide the matched prompt, seed, number of attempts, rejected outputs, or common scoring rubric needed for this comparison.

This research pass inspected the accessible text and provenance around those examples; it did not download and visually score the image pairs. A creator should therefore use them to understand intended range, then make a fresh matched proof for the actual deliverable. Save every candidate, the prompt, ratio, model label, run date, generation ID, and selection reason. Do not describe a curated gallery image as representative performance, and do not infer that a vendor's direct example used the same settings currently exposed by Studio MCP.

Continue with the next operational question

Open the model directory to refresh availability and costs immediately before a run. Use the FLUX-versus-FLUX page when the decision is actually fast iteration versus higher-resolution FLUX output, or the endpoint-identifier guide when automation must distinguish a Studio model ID from a provider endpoint. These links preserve this page's narrow job: selecting between FLUX Pro Ultra and Recraft V3 for one creator-graphics brief.

Required evidence and publication boundary

The current public Studio catalog supports both model labels, their text-to-image workflow, their “best for” positioning, and their 8- versus 5-credit costs. First-party provider documentation supports FLUX Ultra's direct feature and price claims and Recraft V3's direct raster, vector, style, and price claims. The published MCP package supports the narrower Studio discovery and generation contract. Each product fact on this page is cited where it appears.

The required real-sample evidence remains incomplete. Public pages identify attributed examples, but this pass did not establish a controlled, same-prompt Studio set or perform a visual audit. Recent community coverage was also insufficient and degraded. No paid generation, repeated-run test, latency measurement, typography accuracy test, blind panel, or campaign-outcome study was performed. Under the overlap guard, keep actionable model selection in `/models/mcp-model-guides`; this leaf remains draft, unreviewed, researched, and unpublished and offers only an evidence-gap record and future test protocol.