Head to head: Ernie Image Turbo vs Grok Imagine Image 2.0
Ernie Image Turbo vs Grok Imagine Image 2.0
One model showed flashes of taste; the other actually followed the brief. Across eight image-generation tests, this matchup wasn’t competitive.
Ernie Image Turbo never really got out of the gate. On the aggregate, it posts **53.4** to Grok Imagine Image 2.0’s **73.1**, and the task ledger is even harsher: **8 wins to 0**, with **100% confidence** in the result. That’s not a narrow edge or a judge-preference artifact. It’s a clean sweep. What decided it was prompt discipline. Grok repeatedly nailed the things image models usually fumble: exact counts, negation, restricted palettes, and style specificity. It delivered **exactly seven espresso cups**, **exactly 13 relics**, a reading nook with **no visible plants, lamps, or artwork**, and a flat-vector campsite that stayed much closer to the requested four-color constraint. Ernie, by contrast, kept drifting—adding extras, slipping in gradients, or landing near the prompt rather than on it. The gap was just as obvious on more imaginative briefs. In **Mothwing palette hangar**, Grok gave the judges an actual orbital refueling scene with a dragonfly-shaped scout craft, three slender drones, and a tightly controlled palette; Ernie substituted humanoid attendants and leaked stray blue-green tones. In **Neo-Toltec star chart**, Grok understood the assignment as a symbolic mineral-pigment mural with flattened perspective and exactly two astronomer-priests. Ernie made something striking, but it read like polished 3D fantasy art, not the requested visual language. Even where Ernie looked good, Grok was more reliable. The underwater **Clockwork kelp archivist** scene had richer microtexture and a more convincing submerged vault atmosphere in Grok’s version. The **Hands & anatomy** test was worse for Ernie: an apparent extra hand is the kind of mistake that immediately disqualifies an otherwise decent image. That pattern defines the whole matchup—Ernie can produce attractive pictures, but Grok is the model that actually keeps faith with the brief. **Final call: Grok Imagine Image 2.0 wins decisively. Ernie Image Turbo showed occasional visual flair, but Grok was better on adherence, counting, composition, style control, and anatomical plausibility—and it won every single test.**
Mothwing palette hangar
Inside a silent orbital hangar, a dragonfly-shaped scout craft is being refueled by three slender maintenance drones beneath a wide skylight, composed as a clean cinematic 16:9 scene and restricted strictly to the named Mothwing palette: lunar ivory, dust mauve, soot plum, husk brown, and muted lichen gray, with no other stray colors anywhere in the image.
Model B adheres more closely to the prompt by clearly depicting an orbital hangar under a wide skylight, a dragonfly-shaped scout craft actively being refueled by three slender maintenance drones, and a tightly controlled mothwing palette. Model A has a strong cinematic 16:9 composition, but it reads less clearly as orbital refueling, the drones look more like humanoid attendants, and there are noticeable stray cool blue-green tones outside the specified palette. (Second judge pass, order swapped — scores are the average of both: Model B adheres much more closely to the prompt with a silent orbital hangar, a clearly dragonfly-shaped scout craft, three slender maintenance drones actively refueling it, and a palette that stays tightly within the specified mothwing tones. Model A has a strong widescreen composition, but it introduces obvious blue-green stray colors, uses humanoid figures rather than slender maintenance drones, and feels less orbital and less specifically aligned to the named palette.)
Neo-Toltec star chart
A moonlit observatory carved into a basalt cliff above a bioluminescent desert, where two astronomer-priests in feathered copper robes calibrate a brass armillary around a floating turquoise planet, rendered faithfully in a named Neo-Toltec mural style with geometric glyph bands, flattened symbolic perspective, mineral pigments, and crisp stone-texture contours; dramatic side lighting from blue fire braziers, panoramic 16:9 composition.
Model B adheres much more closely to the prompt by depicting exactly two astronomer-priests, a brass armillary around a turquoise planet, geometric glyph bands, flattened symbolic perspective, mineral-pigment mural styling, and the moonlit basalt-cliff observatory above a bioluminescent desert. Model A is visually striking but misses the named mural style and symbolic flatness, adds extra figures, and feels more like a polished 3D fantasy scene than a Neo-Toltec star chart. (Second judge pass, order swapped — scores are the average of both: Model B adheres closely to the Neo-Toltec mural brief, with geometric glyph bands, flattened symbolic perspective, mineral-pigment stone texture, two astronomer-priests, blue fire braziers, and a brass armillary around a floating turquoise planet. Model A is visually striking but departs strongly from the requested mural style and composition, adding extra figures, a photoreal 3D cave look, and an Earth-like globe instead of the specified symbolic panoramic mural treatment.)
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.
Model B adheres much better to the flat-vector campsite brief, with a richer and more balanced camping scene and cleaner graphic styling. Model A has a pleasing wide composition, but it violates the no-gradients restriction more noticeably and feels sparse, with the campsite reduced to a single tent rather than a fuller mountain campsite. (Second judge pass, order swapped — scores are the average of both: Model B closely matches the flat-vector campsite brief with a balanced 16:9 composition and a palette that stays very near the requested four colors, though a few tiny dark specks and subtle tonal variations slightly weaken strict adherence. Model A is visually pleasing but departs more from the prompt by using obvious gradients, a simpler non-mountain campsite scene, and multiple off-palette tonal blends that break the restricted flat-vector requirement.)
Exact counting
A clean overhead flat lay on white marble of EXACTLY seven identical espresso cups arranged in a single straight row, each clearly separate and individually countable, even soft lighting, 16:9.
Model B adheres more precisely to the prompt with exactly seven clearly separate identical espresso cups in a cleaner straight row on white marble, and it presents a more polished overhead flat lay. Model A also has seven cups, but they appear empty rather than espresso cups and the row/alignment feels less refined and less aesthetically strong. (Second judge pass, order swapped — scores are the average of both: Model B matches the prompt very closely with exactly seven identical espresso cups filled with espresso, clearly separated in a straight row on white marble under soft lighting. Model A also has seven cups in a row on marble, but they appear empty rather than espresso cups and the spacing/handling makes the arrangement feel less clean and individually distinct.)
Clockwork kelp archivist
A deep-sea archivist automaton kneels in a submerged library vault, unrolling a sealed map beside shelves encrusted with shells and coral while long ribbons of kelp drift past cracked glass domes; emphasize exceptionally fine texture and detail in the hammered brass joints, braided cable tendons, pitted stone, silted book cloth, filamentous algae, and tiny barnacles, all rendered razor-crisp under narrow shafts of green-blue caustic light.
Model B adheres more closely to the prompt by showing a deep-sea archivist automaton kneeling in a submerged library vault with a sealed map, abundant drifting kelp, cracked glass domes, and much richer microtexture on brass, cables, stone, books, algae, and barnacles. Model A is appealing and readable, but it feels cleaner and less submerged, with fewer of the specified fine details and a less convincing sealed-map/vault atmosphere. (Second judge pass, order swapped — scores are the average of both: Model B better matches the prompt’s deep-sea archivist mood and material specificity, with richer hammered brass, braided tendons, encrusted shelves, drifting kelp, cracked domes, and razor-crisp microtexture under convincing green-blue shafts of light. Model A has a strong, readable composition and appealing lighting, but its robot is cleaner and less intricately detailed, and the scene feels less densely textured and less immersed in the silted, barnacled underwater vault described.)
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.
Model B better matches the prompt with a natural candid photo of one person tying a friendship bracelet, both hands clearly visible, believable finger anatomy, and soft daylight with pleasing shallow depth of field. Model A has strong detail, but it appears to show three hands rather than both hands of one person, which hurts prompt adherence and anatomical plausibility despite decent composition and texture. (Second judge pass, order swapped — scores are the average of both: Model B better matches the prompt with a natural candid feel, both hands clearly visible, and more believable finger anatomy and grip while tying a friendship bracelet in soft daylight with shallow depth of field. Model A has decent focus and texture, but the scene reads less naturally, includes an extra hand, and the anatomy/grip feels less plausible for the requested candid photo.)
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.
Model B adheres more cleanly to the prompt, showing a cozy reading nook with armchair, books, tea, warm afternoon light, and no visible plants, lamps, or artwork. Model A is attractive and cozy, but the composition is tighter and the steaming mug reads a bit more like coffee, while Model B feels more balanced and polished overall. (Second judge pass, order swapped — scores are the average of both: Model B adheres more cleanly to the negation constraints while still delivering a cozy reading nook with warm afternoon light and a balanced 16:9 composition. Model A is also attractive and on-prompt overall, but the visible text on the book spines is a minor distraction and its composition feels a bit tighter and less cleanly arranged than Model B's.)
Thirteen relics on velvet
A top-down still life on a black velvet table in a wizard-cartographer’s cabin, lit by one warm lantern from the upper left, showing EXACTLY 13 distinct relics arranged with clear separation so each is individually countable: 1 obsidian astrolabe, 1 folded silver sextant, 1 glass vial of red sand, 1 petrified seahorse, 1 brass key shaped like a fern, 1 ivory domino engraved with a moon, 1 tiny cobalt idol, 1 coiled copper ring, 1 wax-sealed letter, 1 star map tile, 1 bone whistle, 1 cracked compass, and 1 amber die.
Model B adheres much more closely to the prompt by presenting exactly 13 clearly separated relics and matching nearly all specified items, while Model A includes duplicates and extras that break the exact-count requirement. Model A is attractive and well lit, but Model B has the stronger overall composition for countability and prompt fidelity with similarly strong aesthetics and technical finish. (Second judge pass, order swapped — scores are the average of both: Model B closely matches the prompt with exactly 13 clearly separated relics and nearly all specified items correctly depicted, while Model A includes extra objects and duplicates, breaking the exact-count requirement despite a pleasing layout. Model B is also more coherent as a top-down still life on black velvet with strong lantern lighting and cleaner object readability.)
Matchup powered by OpenRouter.