Last updated Nov 20, 2026

TokenDH same-prompt test: Flash saves 39% vs the flagship

ModelBrief · Published 2026-11-13 · Fact-check 2026-11-18 · Reference: TokenDH

Five days after Flash launched, the first independent same-prompt comparison appeared. On November 13, 2026, TokenDH ran the flagship Seedance 2.5 and Flash on the same prompt in LibTV: a 30-second 720p clip cost 1,380 credits on the flagship and 840 on Flash — about 39% less — with the two versions looking close; the four other Flash clips in the test all shipped on the first try. This article rewrites the test's conclusions and its limits — the limits first: one person, one run per group; LibTV's credit prices are platform promo prices, not to be mixed with Volcano Engine's official ¥0.99/s.

The same-prompt comparison: one scene, two tiers, one run each

DimensionSeedance 2.5 FlashSeedance 2.5 (flagship)
PromptSame prompt: a Song-dynasty warrior entering a city, armed (「持械入宋」)
Length / resolution30 s / 720p30 s / 720p
Platform credits spent840 credits1,380 credits
Saving540 credits, about 39%
ResultThe two versions look close

This is the first test since launch to put the two tiers on the same prompt. The 39% figure points the same way as the official price gap (720p ¥0.99 vs ¥1.51, about 34%) and runs a little larger — but note that credits are LibTV's own billing unit; 840 vs 1,380 is a platform bill, not the official API price.

The other four clips: Flash shipped them on the first try

Besides the comparison, the author ran four more subjects on Flash alone, all usable on the first generation: a 15-second close-combat fight scene — action continuity close to the flagship, good enough to ship; a 15-second Pixar-style 3D animation — finished in one shot; a 30-second family drama with Chinese dialogue — micro-expressions and voice as requested, finished in one shot; and an 8-second osmanthus-fragrance e-commerce clip — product and layout stable, finished in one shot. Across action, animation, dialogue and e-commerce, Flash needed no rework on any of the four.

The author's workflow advice

The author's conclusion is practical: use Flash to test prompts and shots cheaply, then decide whether to re-render on the expensive flagship — and Flash's own output is already usable as-is, not just a "trial tier." That matches this site's buying advice (Flash for drafts and batches, the flagship for finals and high resolution); the difference now is that the sentence has an independent test behind it for the first time.

Limits and pricing caveats

First, the sample is tiny: one person, one run per group, no repeats, so the run-to-run variance is unknown. Second, credits are not the official price: LibTV's credit prices are platform promo prices (720p as low as 0.10 yuan/s during the Double-11 period) — do not convert 840/1,380 credits back into the official ¥0.99/s, and do not write 39% as the official price gap; that gap is about 34%. Third, speed was not timed: the vendor's "near-2×" speed claim was not measured here and stays unverified. Fourth, the subjects are narrow: all five clips are text-to-video, with no image-to-video and no editing or continuation cases.

FAQ

Is Flash 39% cheaper than the flagship?

In this one LibTV comparison, yes: a 30s 720p clip cost 840 credits on Flash vs 1,380 on the flagship, about 39% less. The official API gap is about 34% (720p ¥0.99 vs ¥1.51) — the two figures use different bases, so do not mix them.

Is Flash's output usable as-is?

In this test, yes: all five Flash outputs shipped on the first try, including a 30-second Chinese-dialogue family drama and an 8-second e-commerce clip. But the sample is one person, one run each, so the conclusion is not yet robust.

Does this prove Flash is "closest to Seedance 2.5"?

Only partly: the two versions looked close on the same prompt, but that is one comparison and one run. The vendor's "closest to Seedance 2.5" wording still needs more independent comparisons to verify.

Can LibTV's credit prices convert to the official price?

No. LibTV credit prices are platform promo prices, on a different billing basis from Volcano Ark's ¥0.99/s. Any citation of 840/1,380 must say they are LibTV credits.

ModelBrief's take: This is the first independent test of Flash with a conclusion: the saving is real (about 39% in credits) and the usability is real (five clips shipped on the first try). But it cannot answer the two key questions — how big the quality gap really is (one comparison, one run) and how much faster it is (no timing). Flash's "worth it" question now has its first puzzle piece, not the full picture.

Reference: TokenDH, Nov 13, 2026, original article (rewritten and organized by ModelBrief; conclusions and limits are stated; LibTV credit prices are platform promo prices, on a different basis from the official price)