Replicate FLUX 1.1 pro, Recraft V3 and Ideogram V3 prices vs Sume rows
Replicate lists FLUX 1.1 pro and Recraft V3 at $0.04 and Ideogram V3 quality at $0.09. Sume rows: $0.0375, $0.05, $0.075. 100 images each, compared.

Replicate's pricing page lists black-forest-labs/flux-1.1-pro at $0.04 per output image, recraft-ai/recraft-v3 at $0.04 and ideogram-ai/ideogram-v3-quality at $0.09. Sume does not list those exact versions. Its nearest image rows are FLUX.2 Pro at $0.0375, Recraft V4 at $0.05 and Ideogram V3 at $0.075. For 100 images that is $4.00, $4.00 and $9.00 on Replicate against $3.75, $5.00 and $7.50 on Sume. Source: Replicate pricing (read 2026-10-04).
Note the version differences. FLUX 1.1 pro versus FLUX.2 Pro and Recraft V3 versus Recraft V4 are different models, so this is a price table for planning, not a like-for-like benchmark.
The numbers
Replicate bills these featured models per output image. Sume bills provider list times 1.25 per image, and its Image API docs state that the price shown on an endpoint is the amount charged to your wallet. The Sume ids are black-forest-labs/flux.2-pro, recraft/recraft-v4 and ideogram/ideogram-v3 in the Image API.
| Family | Replicate model | Replicate | Nearest Sume row | Sume | 100 images (R / S) |
|---|---|---|---|---|---|
| FLUX | flux-1.1-pro | $0.04 | FLUX.2 Pro | $0.0375 | $4.00 / $3.75 |
| Recraft | recraft-v3 | $0.04 | Recraft V4 | $0.05 | $4.00 / $5.00 |
| Ideogram | ideogram-v3-quality | $0.09 | Ideogram V3 | $0.075 | $9.00 / $7.50 |
Reading the Ideogram line
Replicate's line is the quality tier by name. Sume's Ideogram V3 row is a single price per image, $0.075, and the quality tiers map onto it in a way described in the catalog. If you rely on the V3 quality tier specifically, run ten prompts on each side and compare before assuming the $1.50 gap per 100 images is free money.
If you are weighing the newer Ideogram, Sume also lists Ideogram 4.5, whose price depends on quality ($0.03, $0.06 or $0.22 per image at fal's list, before Sume's multiplier per the Image API docs).
What changes on a real job
- Retakes: if one in four images is rejected, the effective price is 1.33 times the list. At $0.04 that is about $0.053 per kept image; at $0.0375 it is $0.05.
- References: not every Sume row takes reference images. Check the row's accepted inputs before you plan a reference-based set.
- Failures: Sume does not bill failed or cancelled generations, per the Image API docs. Check Replicate's terms for its own rule before you compare effective costs.
Pick by job, not by list price
A $0.0025 gap per image only matters at volume. At 10,000 images the FLUX line differs by $25. The decision usually turns on the model's look, text rendering and reference support. Use this table to size the budget, then run a 12-image blind test on each candidate and write the winner's price and date into your plan.
Sources
Related posts
More in Comparisons
- Prediction deadlines vs a Sume client deadline: stopping isn't cancel
A client-side deadline stops you watching a Sume job; it does not cancel it. Cancel works only before generation starts, so a missed deadline still bills.
- Runway enterprise exception requests vs Sume's queue-first admission
Runway's docs mention enterprise exception requests for higher volume. Sume takes extra jobs as queued and sets concurrency by plan. How to plan a big batch.
- Runway Characters (live) vs Sume Avatar Video (async)
Runway's API lists gwm1_avatars for real-time conversation. Sume talking faces are async Avatar Video jobs of 4-60 seconds, not live sessions.
- Runway Edit Studio 30 s edits vs Sume's Omni edit
Runway's Aleph 2.0 edits up to 30 s at 1080p in Edit Studio. Sume edits video with gemini-omni-flash-1.1 and a video_url. Compare them, with an API call.
Written by Sume