Loading video...

Video Failed to Load

Go Home

DeepSeek V4.1 Flash vs Gemini 3.8 Flash at 3D geometry test: six self-contained three.js stereometry scenes — a (3,8) torus knot, a Klein bottle, a great stellated dodecahedron, a hyperboloid ruled by 120 straight rods, a level-3 Menger sponge, a thick Möbius strip. one shot each, no libraries beyond...

39,932 views • 1 month ago •via X (Twitter)

14 Comments

Hugo Plat's profile picture
Hugo Plat1 month ago

interesting benchmark

AI/ML API's profile picture
AI/ML API1 month ago

thanks

Tere Labs's profile picture
Tere Labs1 month ago

@cryptoninjap2e hm, DeepSeek is losing this one

AI/ML API's profile picture
AI/ML API1 month ago

@cryptoninjap2e try it on AI/ML API!

Addy Crezee | thehype. | /function1's profile picture
Addy Crezee | thehype. | /function11 month ago

@tere_labs niceeeee

AI/ML API's profile picture
AI/ML API1 month ago

@tere_labs thanks Crezee

AI/ML API's profile picture
AI/ML API1 month ago

our prompts 1) Create a complete, single self-contained HTML file using three.js (everything fully procedural - no textures, no models, no external assets). It renders a mathematically exact (3,8) TORUS KNOT - a rope that winds 3 times around the torus axis and 8 times through its hole - rendered as a thick, perfectly smooth glossy tube. Build it yourself with a BufferGeometry: sample the curve p(u) = ((2 + cos(8u/3)) * cos(u), sin(8u/3) * 1.0, (2 + cos(8u/3)) * sin(u)) for u in [0, 6*PI] with at least 900 samples, compute a stable Frenet-like frame (use a parallel-transport frame so the tube does not twist wildly), and extrude a circular cross-section of radius 0.42 with 48 radial segments around each sample, welding the ends so the tube is a seamless closed loop. The hue runs along the length of the knot so the full spectrum wraps around the whole rope several times and visibly flows along it like current through a cable. 2) Create a complete, single self-contained HTML file using three.js (everything fully procedural - no textures, no models, no external assets). It renders a KLEIN BOTTLE - the classic self-intersecting "bottle whose neck passes through its own wall" figure-8 immersion - as one continuous glossy skin. Build it with a BufferGeometry over the standard parametrization (u in [0, 2PI], v in [0, 2PI], at least 200x200 segments): the bulbous body, the narrowing neck that curves over and PLUNGES THROUGH the side of the body, and the flared mouth that merges smoothly back into the bottom - the surface must be visibly ONE piece with no boundary edge. Render with side: THREE.DoubleSide so the inside of the neck and mouth are visible, and make sure the self-intersection reads clearly. The hue is driven by the u parameter so the spectrum travels along the whole tube from body through neck to mouth, flowing continuously as it animates. 3) Create a complete, single self-contained HTML file using three.js (everything fully procedural - no textures, no models, no external assets). It renders a GREAT STELLATED DODECAHEDRON - the spiky Kepler-Poinsot star polyhedron: 12 regular pentagrammic faces, 20 razor-sharp three-sided spikes radiating outward from an icosahedral core in perfect symmetry, like a crystalline star. Construct it exactly: start from the 12 vertices of a regular icosahedron ((0, ±1, ±phi) and cyclic permutations, phi = 1.618...), build the 20 triangular faces, and raise a sharp pyramidal spike over each face with an apex at 2.6 times the face centroid distance, so the result is a dense sea of long needle-sharp points with clean straight edges and crisp facets. Use flatShading TRUE for this one so every facet is a distinct sharp plane, and give each vertex a hue based on its direction (for example hue from the vertex's spherical angle) so neighbouring spikes glow in different spectrum colors and the whole rainbow rotates around the star as it spins - each spike tip catches a bright white specular flash. 4) Create a complete, single self-contained HTML file using three.js (everything fully procedural - no textures, no models, no external assets). It renders a HYPERBOLOID OF ONE SHEET built as a RULED SURFACE out of individual straight rods - the classic "impossible" curved shape made only from straight lines, like a cooling tower or a hyperboloid gear model. Create 120 thin glossy cylinders (radius 0.055, length spanning the full height): take 120 points evenly spaced on a circle of radius 2 at height y = -3, and connect each to a point on the SAME radius circle at y = +3 that is rotated by a fixed twist angle of 75 degrees - each rod is perfectly straight, yet together they sweep out the smooth hourglass waist. Add two thin glossy rings (torus meshes) capping the top and bottom circles. Each rod gets its own hue from its angular position so the 120 rods form a continuous rainbow ring, and the hue offset shifts every frame so the colors race around the hyperboloid like a spinning spectrum. ALSO animate the twist angle very subtly between 70 and 80 degrees with a period of exactly 5 seconds (rebuild or re-orient the rods) so the waist gently pulses - the loop must stay perfectly seamless. 5) Create a complete, single self-contained HTML file using three.js (everything fully procedural - no textures, no models, no external assets). It renders a MENGER SPONGE fractal cube at RECURSION LEVEL 3 - a cube riddled with square holes at three nested scales (20 sub-cubes per level, 8000 small cubes total): start from a cube of side 6, recursively subdivide each cube into a 3x3x3 grid and delete the 6 face centers and the 1 body center, three levels deep. For performance use a SINGLE THREE.InstancedMesh (or a merged BufferGeometry) for all 8000 cubelets, never 8000 separate meshes. Every cubelet is a crisp glossy cube with slightly beveled-looking shading. The hue of each cubelet is a function of its 3D position (for example hue = fract((x + y + z) * 0.12 + time)) so the rainbow sweeps DIAGONALLY through the whole sponge in waves - set per-instance colors via instanceColor and update them each frame so the spectrum visibly flows through the fractal. The intricate holes must read clearly: keep the object bright and let the light reach into the tunnels. 6) Create a complete, single self-contained HTML file using three.js (everything fully procedural - no textures, no models, no external assets). It renders a MOBIUS STRIP with real thickness - a wide ribbon with a single half-twist, so it has only ONE side and ONE edge, rendered as a solid glossy band you could hold. Build it with a BufferGeometry over u in [0, 2*PI] (at least 400 segments) and v in [-1, 1] (at least 40 segments): center circle radius 3, ribbon half-width 1.0, one half-twist (the cross-section rotates by u/2), and give the ribbon a real thickness of 0.18 by generating both an outer and an inner surface plus the closed side edges, so the band is a watertight solid with a visible cut-free edge. The hue runs along the u parameter so the spectrum wraps the entire band exactly twice (proving the one-sided topology as the color returns to itself only after two laps) and flows continuously along the ribbon while it rotates.

Jeudry Peña Peña's profile picture
Jeudry Peña Peña1 month ago

How is gemini 3.8 flash cheaper than deepseek??

AI/ML API's profile picture
AI/ML API1 month ago

deepseek isn't more expensive per token — it just reasons a lot longer before writing code. ~168k output tokens total vs gemini's ~21k for the same 6 scenes, that's what drove the cost.

AshutoshShrivastava's profile picture
AshutoshShrivastava1 month ago

DeepSeek dropping banger after banger.

AI/ML API's profile picture
AI/ML API1 month ago

yeah but it costs more :(

AshutoshShrivastava's profile picture
AshutoshShrivastava1 month ago

outputs is also better

Jigs's profile picture
Jigs1 month ago

Not everytime DeepSeek

AI/ML API's profile picture
AI/ML API1 month ago

yes, in this case Gemini is cheaper and (in my opinion) better even despite a bug in the first scene

Related Videos

glm 5.3 vs qwen 3.8 vs gemini 3.7 vs deepseek v4 flash four models designed and built three structures each on a physics-backed site, with no dimensions anywhere in the brief the setup: our own agent loop on OpenRouter, a construction site as the tool set – footings, walls, arches, roofs, scaffold, a lamp. the site enforces physics and nothing else: unsupported brick falls, a roof needs walls under it, a worker reaches 3.2 m above whatever he stands on, an arch needs centring until the keystone is set, concrete cures before it carries. no budget ceiling – material cost is tallied and reported, never blocked. tasks: 1. house – a plot and a palette, no plan. shape, height and material are the model's call 2. lighthouse – a headland cut by a gully, with a rock stack standing 30 m offshore. the lamp must burn, it must be the highest thing built, and the keeper must be able to walk to it 3. bridge – a river with one islet and banks at different heights. cross it however you want models: Z.ai glm 5.3 flash, Qwen qwen 3.8 flash, Google DeepMind gemini 3.7 flash, DeepSeek v4 flash vision all twelve objects were finished and signed off by the models themselves. tallest lighthouse is qwen's at 38.4 m, planted on the offshore stack with a bridge run out to it – the only model that read the site that way. deepseek signed off its bridge on an empty riverbed: 0 bricks, 107 minutes, $1.16m of material tallied - total cost, three builds #1 glm 5.3 flash – $0.201 #2 gemini 3.7 flash – $0.871 #3 qwen 3.8 flash – $1.058 #4 deepseek v4 flash – $1.567 - wall clock, three builds #1 gemini 3.7 flash – 91m #2 glm 5.3 flash – 228m #3 deepseek v4 flash – 502m #4 qwen 3.8 flash – 912m - total tokens #1 gemini 3.7 flash – 3,567,052 #2 glm 5.3 flash – 4,732,748 #3 qwen 3.8 flash – 13,469,333 #4 deepseek v4 flash – 18,230,076 - defects logged by the site #1 deepseek v4 flash – 59 #2 gemini 3.7 flash – 132 #3 glm 5.3 flash – 221 #4 qwen 3.8 flash – 350 - material tallied across three builds #1 gemini 3.7 flash – $359,884 #2 glm 5.3 flash – $583,358 #3 deepseek v4 flash – $1,327,484 #4 qwen 3.8 flash – $2,188,625 observations: • glm is the cheap one and nothing here is close – $0.201 for three buildings, $0.042 per million tokens, 6x under gemini's rate • what glm spends it on is bulk, not care: 166,228 bricks in one house and 156 defect weight, the worst single object in the set • gemini is the efficiency line – 91 minutes and 3.57m tokens for all three and an eighth of qwen's clock • gemini also builds the smallest of everything. its lighthouse is 22.5 m against qwen's 38.4, its house 6.9 m against 19.3 • qwen is the maximalist: 1.18m bricks, $2.19m of material, tallest on all three tasks, and 912 minutes – 15 hours – to get there conclusion: twelve finished objects for $3.80 all in, and a 7.8x price spread between the cheapest model and the priciest! follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

26,360 views • 2 months ago

gemini 3.7 flash vs deepseek v4 pro 0813 vs muse spark 1.2 – on voxel city dioramas three models each built three crossy road-style 3d scenes – a construction site, a nyc intersection, a river with a drawbridge – as single self-contained html files the setup: Nous Research's hermes agent cli on OpenRouter, three.js skills preloaded, identical prompts per scene tasks: 1. construction site – tower crane on a working lift loop, paver laying fresh road, roller compacting it behind 2. nyc crossing – four-way intersection with a traffic light state machine, queuing cars, pedestrians crossing on the walk signal 3. river drawbridge – double-leaf bascule that lifts for tall boats, cars queuing at the barriers, animated water every scene: Three.js r185, box geometry only, a locked 20-color palette, four camera presets, and a day/night mode with bloom. one file, no build step, no assets models: Google DeepMind gemini 3.7 flash, DeepSeek v4 pro 0813, AI at Meta muse spark 1.2 muse and gemini finished every scene in two to three minutes. deepseek took 15 to 41 minutes per scene - build time, all three scenes #1 gemini 3.7 flash – 6m 43s #2 muse spark 1.2 – 7m 20s #3 deepseek v4 pro – 91m 25s - total tokens #1 muse spark 1.2 – 440,279 #2 gemini 3.7 flash – 713,855 #3 deepseek v4 pro – 20,957,568 - total price #1 muse spark 1.2 – $0.53 #2 gemini 3.7 flash – $0.56 #3 deepseek v4 pro – $4.57 - agent calls across the three builds #1 muse spark 1.2 – 12 #2 gemini 3.7 flash – 18 #3 deepseek v4 pro – 143 observations: • muse won two of the three scenes on looks with the smallest files in the test – 887 to 1,042 lines against gemini's 1,934 to 2,377. cheapest, fastest to a good frame, and shortest turned out to be the same column • deepseek burned 20.96m tokens – 29x gemini, 48x muse – across 143 agent calls. prompt caching is the only reason that cost $4.57: the cache discount absorbed roughly $30 of resent context • gemini was the only model whose files needed zero fixes to render – and the only one whose night mode is cosmetic. the sky never darkens and one camera button does nothing. clean code for a scene it never looked at follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

29,033 views • 2 months ago

ox alpha vs deepseek v4 flash vision vs grok 4.6 vs gemini 3.7 flash vs – on photo-to-3d four vision models got one photograph each and had to rebuild the place inside it as a Three.js scene. twelve scenes, twelve first-try runs, zero console errors the setup: one reference photo per scene, sent as an image on OpenRouter. the prompt never says what is in the picture – no "motel", no "bar", no "gas station". the model has to read the photo and rebuild it: layout, materials, hour of the day, and whatever is around the corner that the frame does not show tasks – three photographs of early-2000s america: 1. a motel at night, neon pylon lit, snow on the ground 2. an old new york tavern interior, tin ceiling, tiled floor 3. an abandoned service station in the california desert, midday sun each scene ships as one self-contained html file, procedural geometry and canvas textures only, no downloads. three timed camera shots, and shot 1 has to reproduce the framing of the reference photo models: xAI grok 4.6, Google DeepMind gemini 3.7 flash, DeepSeek deepseek v4 flash vision exp, and ox alpha – a stealth model on openrouter, free, no lab attached to it yet results: - wall clock, three scenes #1 gemini 3.7 flash – 11m 12s #2 deepseek v4 flash – 15m 20s #3 grok 4.6 – 28m 11s #4 ox alpha – 38m 54s - output tokens #1 gemini 3.7 flash – 77,396 #2 ox alpha – 87,613 #3 grok 4.6 – 105,687 #4 deepseek v4 flash – 127,884 - lines of code shipped #1 ox alpha – 2,090 #2 deepseek v4 flash – 2,291 #3 grok 4.6 – 3,529 #4 gemini 3.7 flash – 3,989 - total price #1 ox alpha – $0.000 #2 deepseek v4 flash – $0.091 #3 gemini 3.7 flash – $0.136 #4 grok 4.6 – $0.697 observations: • grok is 7.7x the price of deepseek. it is the only model that read the light – low sun, real shadows on the station, a cold night on the motel • gemini is the fastest and the least deliberate. 17,158 reasoning tokens against deepseek's 99,172, and it still shipped the most code – 3,989 lines • deepseek thought hardest and rendered plainest. 99,172 reasoning tokens, 5.8x gemini's, spent on layout rather than on light. its motel is the second best in the set for $0.030 • ox alpha is free and reads a photo as well as anything here – it lifted "family units / kitchenettes" off the pylon and redrew it in canvas conclusion: twelve scenes, four models, zero fixes, and the whole run cost $0.924! follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

24,470 views • 2 months ago