Skip to main content

Direct comparison

FLUX.3 Video vs. Sora 2

Compare the exact model versions across shared blind-preference benchmarks, public pricing, release dates, resolution, duration, native audio, access, and model availability.

Overview

Technical specifications and pricing

Technical limits come from model or API documentation. A representative minute price appears only when it can be traced to the exact model. Different resolutions and audio settings may still affect the final charge.

Specifications and pricing for FLUX.3 Video and Sora 2
AttributeFLUX.3 VideoSora 2
ProviderBlack Forest LabsOpenAI
ReleasedAugust 4, 2026September 2025
AccessAPI, Web appAPI
WeightsProprietaryProprietary
ParametersNot publishedNot published
Maximum resolutionNot published720p
Maximum durationNot published20 s
Native audioYesYes
Representative API priceNot published$6 per minute Source
LifecycleNo separate noticeMarked as legacy. API shutdown announced for September 24, 2026 Status source
Technical sourceModel sourceVerified 2026-08-24Model sourceVerified 2026-08-24

Performance

Shared video benchmarks

A row appears only when both exact model versions occur in the same LMArena task and stored dataset revision. Text-to-video and image-to-video are never merged into one score.

FLUX.3 VideoSora 2
LMArena Text to VideoBlind preference rating from the same task and dataset revision
FLUX.3 VideoHigher value1,494.4
Sora 21,340.1

Key to the marks:better valuebetter, but not conclusivetie

Shared video benchmarks for FLUX.3 Video and Sora 2
BenchmarkFLUX.3 VideoSora 2Source
LMArena Text to Video
Higher rating1,494.4Rank 2, 95% interval 1,477.1 to 1,511.8, 1,291 votes
1,340.1Rank 16, 95% interval 1,333.5 to 1,346.7, 62,692 votes
LMArenaRetrieved 2026-08-26

Higher rating is better. A clear lead is shown only when the published 95% intervals do not overlap.

Key to the marks:better valuebetter, but not conclusivetie

Visual comparison

Sample videos

Only runs from both exact model versions under the same archived prompt and first-output rule appear here. Provider showcases and substitute versions remain excluded.

Controlled practical comparison. Both exact model versions received the same archived prompt in one request. The first output counts. Playback uses the first five seconds, 1280 × 720 pixels, 30 fps, no crop, and muted audio. Provider-side randomness and prompt processing remain model characteristics.

Motion and physical consistency

Anatomy, object consistency, camera tracking, rain, and reflections

FLUX.3 Video

Run details

fal.ai. blackforestlabs/flux-3/text-to-video. 16:9, one request, first output, requested 5 s, seed 2955816195, request {"prompt":"Generate exactly one continuous five-second 16:9 cinematic video. A fictional bicycle courier in a mustard-yellow raincoat rides a red city bicycle from left to right across a rain-wet Hamburg street at blue hour for the entire clip. The camera tracks her smoothly at waist height. Both hands remain visible on the handlebars, the bicycle geometry stays physically correct, and a transparent umbrella fixed to the bicycle moves naturally in the wind. Wet-street reflections and raindrops follow realistic physics. No cuts, no slow motion, no text, no logos, no subtitles, and no audio.","aspect_ratio":"16:9","resolution":"720p","duration":5,"generate_audio":false}, configured price estimate $0.85.

Generated Aug 24, 2026. Source

Sora 2

Run details

fal.ai. fal-ai/sora-2/text-to-video. 16:9, one request, first output, requested 8 s, seed provider random, request {"prompt":"Generate exactly one continuous five-second 16:9 cinematic video. A fictional bicycle courier in a mustard-yellow raincoat rides a red city bicycle from left to right across a rain-wet Hamburg street at blue hour for the entire clip. The camera tracks her smoothly at waist height. Both hands remain visible on the handlebars, the bicycle geometry stays physically correct, and a transparent umbrella fixed to the bicycle moves naturally in the wind. Wet-street reflections and raindrops follow realistic physics. No cuts, no slow motion, no text, no logos, no subtitles, and no audio.","resolution":"720p","aspect_ratio":"16:9","duration":8}, configured price estimate $0.80.

Generated Aug 24, 2026. Source

Prompt and integrity hashes

Generate exactly one continuous five-second 16:9 cinematic video. A fictional bicycle courier in a mustard-yellow raincoat rides a red city bicycle from left to right across a rain-wet Hamburg street at blue hour for the entire clip. The camera tracks her smoothly at waist height. Both hands remain visible on the handlebars, the bicycle geometry stays physically correct, and a transparent umbrella fixed to the bicycle moves naturally in the wind. Wet-street reflections and raindrops follow realistic physics. No cuts, no slow motion, no text, no logos, no subtitles, and no audio.

Prompt SHA 256: 206a9182584e93790fd6df608e8a2065fe2c8f747e85aeaaf4c8ea84040e6299

FLUX.3 Video, Original SHA 256: 64c9a4f8661b0bd659d4d4357af0e819cb8606458507270bc8fbb91da81b4a52

FLUX.3 Video, Normalized SHA 256: 23c7205d58bd75b07179a73bd51c27f1eb37522e4037a9c73803c0c4ff12edb9

Sora 2, Original SHA 256: 017b48183183b34289b75843b0d07bf5ba373800adfcc4d148a17979ef322538

Sora 2, Normalized SHA 256: 0b391c3b44e83bcd4deed191bbd65ca924385d2a274a42ff95d79722023c0728

How to read the data

Preference is not the same as production fit

A higher LMArena rating means that outputs from this exact configuration were preferred more often in blind comparisons.

Motion control, identity consistency, audio, turnaround time, moderation, and licensing may matter more than the aggregate preference rating.

Prices are representative public minute rates, not a normalized cost benchmark.

More matchups

Related video model comparisons

Compare FLUX.3 Video and Sora 2 with other important video models using the same benchmark logic.

View the complete video model comparison