Head to head: FLUX 2 vs Ideogram V4.5 Edit
FLUX 2 vs Ideogram V4.5 Edit
By Ryan Merket · Published
FLUX 2 and Ideogram V4.5 Edit trade strengths across a varied set of image-generation and editing prompts. The comparison turns on different kinds of prompt fidelity, not a decisive overall edge.
The saved result is a tie: FLUX 2 takes four task wins, Ideogram V4.5 Edit takes three, and one task is tied. Aggregate scores are **59.5** for FLUX 2 and **61.6** for Ideogram V4.5 Edit, but the saved verdict remains a tie—not an overall win for the model with the higher score. FLUX 2’s best results came when the brief demanded a busy scene or a precise arrangement: it captured more of the Memphis avenue’s named details, the fruit kiosk’s left-side golden light and overhead wires, and the violinist’s intimate, distressed portrait. It also won attribute binding, with the requested objects and their positions clearly legible. Ideogram V4.5 Edit stood out on the rainy bus shelter, where the courier and map-reading man were easier to distinguish, and on the bracelet-making scene, with more convincing hands and working grip. Its flat-vector campsite also adhered more closely to the restricted palette. Reflections and glass was the lone tie: Ideogram put the apple on the counter and showed reflections, while FLUX better matched the dark marble and dramatic lighting. Neither model owns the full brief: FLUX more often preserved specific scene details, while Ideogram was stronger on figure separation, hands, and palette discipline. **Final call: Too close to call (not statistical).** This assessment applies to the saved task sample.
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 B places the red apple on the counter and shows it reflected in the chrome, while also providing clear window reflections and a polished scene. Model A better matches the dark marble and dramatic lighting, but its apple is on the windowsill rather than the counter, weakening adherence to a key detail. (Second judge pass, order swapped — scores are the average of both: Model B more closely follows the prompt, with the apple on the counter and visible window and apple reflections in the chro | Anthropic: Claude Fable 5.1: Model A nails the dark polished marble, dramati…
Memphis laundromat avenue
A bustling urban avenue reimagined in bold 1980s Memphis design style: a corner laundromat named 'Spin Comet 24' with zigzag awnings, squiggle-patterned sidewalks, teal-and-coral sedans, black-and-white geometric window grids, a woman walking three dalmatians past a giant yellow detergent box sign, everything flattened into crisp vector shapes, playful asymmetry, high-contrast pastel palette, stylized shadows, poster-like composition, 16:9
OpenAI: GPT-6 Luna: Model A more fully captures the requested bustling Memphis-style avenue, with the named corner laundromat, zigzag awnings, patterned sidewalk, teal cars, yellow detergent sign, and woman walking dalmatians. Model B has a clean, appealing illustration, but its oversized “SUDS” sign competes with the requested laundromat name and its palette feels less bold and high-contrast. (Second judge pass, order swapped — scores are the average of both: Model A more fully captures the bold Memphis-style aven | Anthropic: Claude Fable 5.1: Model A hits more literal beats (corner shop, z…
Rainy bus shelter materials study
A believable real-world street scene under overcast post-rain light: a city bus shelter on Calle 47 with a brushed aluminum bench, wet cedar slats, a fogged glass ad panel, a courier in a mustard nylon jacket holding a cardboard tube, and an elderly man with dark brown skin and a wool cap reading a folded transit map, every surface rendered with physically convincing material response—beads of water on metal, damp wood grain, diffused reflections in puddles, soft skin sheen, matte fabric fibers—camera at eye level from just outside the shelter, naturalistic documentary photography style, 16:9
OpenAI: GPT-6 Luna: Model B more clearly separates the courier with the cardboard tube from the elderly man reading a folded map, and it convincingly includes the Calle 47 shelter, metal bench, wood slats, and rain-wet pavement. Model A captures the rainy setting and materials, but both figures wear similar yellow jackets and the roles are less distinct. (Second judge pass, order swapped — scores are the average of both: Model B clearly separates the mustard-jacketed courier from the elderly man reading a map, but | Anthropic: Claude Fable 5.1: Model B correctly separates the two characters—a…
Golden-hour fruit repair kiosk
A photorealistic urban side-street scene at golden hour: a narrow cracked pavement in the Mercado de las Nubes district where a tiny kiosk called 'Bruno's Fruit Radio Repair' displays open transistor sets beside pyramids of blood oranges, the low sun skimming in from frame left and turning dust in the air amber, long shadows from overhead tram cables striping the ground, warm rim light catching a mechanic in a teal apron as she solders a radio while a child in silver rain boots watches, mood wistful and hopeful, cinematic composition, 16:9
OpenAI: GPT-6 Luna: Model A more closely matches the requested left-side golden hour, cracked pavement, overhead cables, orange pyramids, teal-aproned mechanic, and child in silver boots. Model B has a more cinematic kiosk composition and includes the district sign, but misses the specified lighting direction and overhead-cable details. (Second judge pass, order swapped — scores are the average of both: Model A more clearly captures the low sun entering from frame left, overhead tram wires, and cracked pavement, wh | Anthropic: Claude Fable 5.1: Model A nails the low sun from frame left, crac…
Missed-call violinist
A close street-portrait of a young violinist standing beside the shuttered entrance of the Lorca Avenue metro stop, convincingly conveying stunned heartbreak after seeing a missed call on his phone: brows slightly lifted and drawn together, lower lip parted, eyes glassy but not crying, one hand still gripping the violin case too tightly, shoulders frozen mid-breath; nighttime urban background softly blurred with a vending machine glow and a single red taxi streak, intimate cinematic framing, photorealistic style
OpenAI: GPT-6 Luna: Model A better matches the intimate street-portrait and includes the violin, case, phone, shuttered entrance, vending-machine glow, and red taxi streak; the expression also reads as distressed. Model B has a polished cinematic look, but its distant full-body framing and lack of a visible violin weaken the prompt match. (Second judge pass, order swapped — scores are the average of both: Model A better captures the intimate portrait and stunned, heartbroken expression, with the violin, phone, vend | Anthropic: Claude Fable 5.1: Model A delivers the requested close, intimate…
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 B shows both hands in a plausible working grip with a visible braided bracelet and crafting materials, while Model A’s fingers and tiny thread interaction look less natural. Model A has soft, shallow-focus lighting, but Model B better captures the candid bracelet-making scene overall. (Second judge pass, order swapped — scores are the average of both: Model B more clearly depicts a friendship bracelet being worked with colorful threads, and its tabletop setting, soft daylight, and shallow | Anthropic: Claude Fable 5.1: Model B delivers a genuinely candid scene with a…
Attribute binding
A photorealistic tabletop scene: a matte green cube to the LEFT of a glossy red sphere, a small blue cylinder BEHIND both, and a single yellow rubber duck sitting ON TOP of the cube. Colors must stay on the correct objects. Soft studio light, 16:9.
OpenAI: GPT-6 Luna: Model A clearly shows the matte green cube on the left, the glossy red sphere to its right, the blue cylinder behind them, and the yellow duck sitting on top of the cube, with polished studio lighting. Model B includes the same objects and relationships, but has more empty space, a noisier finish, and a less glossy-looking sphere. (Second judge pass, order swapped — scores are the average of both: Both images include the correctly colored objects in the requested arrangement, but Model A separat | Anthropic: Claude Fable 5.1: Both bind colors correctly and place the duck o…
Restricted palette
A flat-vector illustration of a mountain campsite at dusk using ONLY four colors — cream, burnt orange, teal, and charcoal — with no gradients and no stray off-palette colors, balanced composition, 16:9.
OpenAI: GPT-6 Luna: Model A has a richer, well-balanced campsite scene, but its gradients and many tonal variations conflict with the strict four-color restriction. Model B is a cleaner flat-vector illustration and stays much closer to the requested palette, though a few details introduce extra shades and its composition is simpler. (Second judge pass, order swapped — scores are the average of both: Model B follows the restricted four-color, flat-vector brief much more closely, though a few gray and brown details s | Anthropic: Claude Fable 5.1: Model A is far more polished and atmospheric bu…
Matchup powered by OpenRouter.