Head to head: Meta Muse Image Edit vs FLUX 2
Meta Muse Image Edit vs FLUX 2
By Ryan Merket · Published
Eight image-editing challenges put hand detail, exact prompt constraints, reflections, named styles and scene composition under the same scrutiny. The results turn on more than visual polish: each image also had to deliver the brief’s specific details.
Meta Muse Image Edit wins across all eight tasks, with an aggregate score of 66.8 to FLUX 2’s 55.8 and a limited confidence verdict. The gap shows up in prompt fidelity as much as finish: Muse included the Kestrel Bay terminal sign and 5:40 a.m. display, and gave the ferry poster its “Moss Gull” name—details FLUX missed or weakened. Muse was also more reliable when composition and subject matter had to work together. Its friendship-bracelet image read as the act of tying, its ukiyo-e scene clearly showed a fishing boat and crew, and its cargo lineup followed the requested diagonal while distinguishing the objects’ materials. FLUX had attractive images of its own, but its bracelet looked closer to an inspection, its boat less clearly fished, and its lineup was mostly horizontal. The advantage held in the more atmospheric scenes and reflective surfaces. Muse made the chrome and marble reflections more convincing, gave the empty terminal richer detail without adding buses, and captured the wet tracks and numbered red funicular at the tram depot. FLUX sometimes had a local edge—the tea steam was more visible, and its canal scene had an effective misty mood—but those touches did not make up for weaker adherence to the full briefs. **Call: Meta Muse Image Edit wins decisively, 8–0.** FLUX 2 can make appealing images, but in this matchup Muse more consistently turned specific instructions into coherent, finished scenes.
Hands & anatomy
A natural candid photo of a person tying a friendship bracelet, both hands clearly visible with anatomically correct fingers and a plausible grip on the threads, soft daylight, shallow depth of field, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly shows a candid moment of tying a friendship bracelet, with both hands working colorful loose threads and a pleasing shallow-focus daylight setting. Model B has a clean close-up, but it reads more like inspecting a mostly finished bracelet than tying one and is less compositionally rich. (Second judge pass, order swapped — scores are the average of both: Model B clearly centers both hands and shows a plausible grip on a formed friendship bracelet, with soft light and shallow | Anthropic: Claude Fable 5.1: Model A delivers a convincing candid scene with…
Negation
A cozy reading nook with an armchair, a stack of books, and a mug of tea by a window — with absolutely NO plants, NO lamps, and NO artwork or picture frames anywhere in the frame. Warm afternoon light, 16:9.
OpenAI: GPT-6 Luna: Model A delivers a bright, polished reading nook with all requested objects and none of the prohibited plants, lamps, or artwork; Model B also follows the constraints and matches the requested 16:9 framing, but its darker, grainier presentation is less inviting. Model A’s composition and warm afternoon light are especially appealing, though Model B makes the tea steam more visible. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested 16:9 frami | Anthropic: Claude Fable 5.1: Both Model A and Model B successfully omit plan…
Reflections & glass
A photorealistic close-up of a chrome teapot on a polished dark marble counter next to a glass of water; the window and a red apple on the counter must be correctly reflected in BOTH the chrome and the water's surface, dramatic side light, 16:9.
OpenAI: GPT-6 Luna: Model A is more photorealistic and compelling, with crisp window reflections in the chrome and polished marble, though the apple reflection in the water is optically unclear. Model B includes the requested objects but has a less convincing teapot shape and muddier reflections, and the apple is not clearly reflected in both surfaces. (Second judge pass, order swapped — scores are the average of both: Model A has more convincing, detailed reflections of the window in the chrome and glass, with str | Anthropic: Claude Fable 5.1: Model A delivers a far more photorealistic, dra…
Named art style
A ukiyo-e woodblock print of a fishing boat riding a large cresting wave at dawn, faithful to the flat color planes, bold outlines, and stylized foam of the tradition, muted indigo and cream palette.
OpenAI: GPT-6 Luna: Model A more clearly depicts a fishing boat and its crew, while its cresting wave, dawn landscape, muted indigo-and-cream palette, and woodblock-like outlines strongly match the prompt. Model B has a clean, striking ukiyo-e composition, but the boat reads more like a larger sailing vessel and is less clearly engaged in fishing. (Second judge pass, order swapped — scores are the average of both: Model A more clearly depicts a fishing boat with its crew riding a dramatic cresting wave, while stron | Anthropic: Claude Fable 5.1: Model A delivers a convincing ukiyo-e print wit…
Bound-color cargo lineup
A studio-clean isometric 3D render of four transport objects arranged in a precise diagonal lineup on a matte charcoal floor: a cobalt blue brushed-steel scooter, a saffron yellow plastic fuel can, an emerald green glass signal lantern, and a crimson red leather pilot helmet, each object clearly separated with its own color and material preserved exactly, soft overhead softbox lighting, sharp shadows, 16:9.
OpenAI: GPT-6 Luna: Model A better follows the requested diagonal lineup and preserves the distinct colors and materials, with especially convincing brushed metal, glass, and leather details. Model B keeps all four objects fully in frame, but its arrangement is mostly horizontal and its objects are less faithful to the requested finishes and pilot helmet. (Second judge pass, order swapped — scores are the average of both: Model B presents all four requested objects clearly separated in the specified colors and mate | Anthropic: Claude Fable 5.1: Model A delivers a true diagonal lineup from an…
Melancholy vector ferry poster
A clean flat vector illustration of a compact river ferry named Moss Gull gliding through a misty canal at dawn, muted teal water, pale apricot sky, simplified geometric ripples, two bundled commuters under a tiny canopy, mood of quiet melancholy and first-light stillness, crisp poster design with generous negative space.
OpenAI: GPT-6 Luna: Model A more closely matches the clean flat-vector poster brief, with the readable “Moss Gull” name, two bundled commuters beneath a canopy, geometric teal ripples, and generous apricot-sky negative space. Model B conveys a misty canal atmosphere effectively, but its more rendered, detailed treatment and lack of the ferry name make it less faithful to the requested design. (Second judge pass, order swapped — scores are the average of both: Model A more closely matches the clean flat-vector poste | Anthropic: Claude Fable 5.1: Model A delivers the prompt's core specifics —…
Empty bus terminal, no buses
An ultra-detailed isometric 3D render of the Kestrel Bay intercity terminal at 5:40 a.m., polished concrete platforms, glowing timetable boards, vending machines, puddles from overnight cleaning, a lone janitor cart and scattered pigeons, but absolutely no buses anywhere in the scene, cool fluorescent lighting, orderly composition, 16:9.
OpenAI: GPT-6 Luna: Model A more completely matches the requested terminal, including the Kestrel Bay signage, 5:40 a.m. display, puddles, janitor cart, pigeons, and no buses, with especially rich detail. Model B has a clear empty platform and the requested supporting elements, but omits the specific terminal and time cues and includes a janitor rather than just the cart. (Second judge pass, order swapped — scores are the average of both: Model B captures the isometric platform scene and includes the janitor, pigeo | Anthropic: Claude Fable 5.1: Model A delivers a richly detailed isometric te…
Golden-hour tram depot
A rain-damp hilltop tram depot in Valparaíso at golden hour, with a cherry-red 1930s funicular car numbered 27 pausing beside corrugated workshops, long amber sunlight grazing the metal roofs and throwing stretched shadows across the tracks, a faint sea haze beyond the ridge, cinematic documentary realism, low three-quarter view, 16:9.
OpenAI: GPT-6 Luna: Model A more fully captures the rain-damp depot, the clearly numbered cherry-red funicular, corrugated workshops, and amber light reflecting across the tracks. Model B has the requested wide framing and strong coastal hillside setting, but feels drier and less like a working depot, with the car and its number less prominent. (Second judge pass, order swapped — scores are the average of both: Model A more vividly conveys the rain-damp depot, hillside setting, workshop details, and golden-hour lig | Anthropic: Claude Fable 5.1: Model A nails the rain-damp tracks, cherry-red…
Matchup powered by OpenRouter.