Fluid
Fluid — WebGL fluid simulation with swirling particles.
What I asked each model — the Fluid prompt
Every model on this page got this exact prompt inside the Agent Operating System: Fluid — WebGL fluid simulation with swirling particles.
Single HTML file out. No iteration. No examples in the system prompt. Whatever each model produced on the first run is what's on this page. 23 frontier models have attempted it so far: Claude Fable 5, Fugu Ultra, Fugu Mini, Fusion, Gemini 3.6 Flash, GLM-5.2, GPT-5.6 Sol, Grok, Inkling, Kimi K2.7, Kimi K3, MiniMax M3, Hermes MoA, Opus 4.8, Claude Opus 5, Qwen 3.8, Qwen 3.7, Claude Sonnet 5, DeepSeek V4 Pro, DeepSeek V4 Flash, Kimi K2.7 · Fast, Kimi K2.7 · No-Think, Kimi K2.7 · Quality.
Why this task matters. Fluid is a textbook test of sim-class capability — the kind of build that exposes whether a model is doing pattern-matching or actual reasoning. Shipping this cleanly is the floor for what I expect from a frontier model — every model on the leaderboard should at least attempt it.
How each model handled Fluid
Ranked by my 0–10 score from the source comparison guides on agentos.guide. Click any to play the actual one-shot HTML the model produced.
What I saw: Renders beautifully with elegant curl-noise flow lines and a lovely cyan-to-magenta speed-based gradient, plus autopilot vortex and clean pointer interaction. Strong and shippable, but it's more a flow-field particle field than a true fluid sim with density diffusion, so it falls just short of topping the best.
What I saw: Ultra v2 — fluid sim. Smoke-test MAYBE (0.0% diff) — needs click-drag injection which generic input didn't replicate.
What I saw: 2D fluid simulation, click-drag density+velocity. Smoke-test PASS.
What I saw: Stable-fluids style 2D simulation with click-and-drag density+velocity injection. Polished glass HUD, gradient title, live pill indicator, right-side controls panel. Touch-action set so it works on mobile. Visually on par with Opus's blob-in-bowl.
What I saw: Strong particle density (75k) with glowing additive blending, polished glassmorphism UI, presets and shockwave controls all render cleanly; but the center is blown out to solid white and it reads more as a bright particle cloud than an elegant swirling fluid, keeping it just short of the field's best.
What I saw: GLM filled the bowl with glowing liquid that actually sloshes — the most convincing 'liquid in a bowl'. Opus's particles glowed but clumped to the centre. Kimi's collapsed into a tiny blob.
What I saw: Gorgeous swirling vortex of multicolored particle trails around a glowing core — the cyan/violet/pink palette, additive-blended trails, and clean UI chrome (title, metrics, custom cursor) make it genuinely polished and clearly on-brief. Not a literal Navier-Stokes fluid sim but the flow-field aesthetic reads as fluid convincingly, and the composition tops the field.
What I saw: A round bowl of ~500 distinct glowing droplets that fall, pool at the bottom and slosh when you tilt with the mouse, with colour themes. The first version melted into one white blob; this one keeps every droplet separate, like real liquid.
What I saw: Renders cleanly with a polished Orbitron title and a dense particle cloud with pleasing color mix, but it reads as a scattered particle sphere rather than convincing swirling fluid motion — the flow field is subtle and the CPU per-particle loop limits it to a generic point cloud rather than the cohesive vortex the brief implies.
What I saw: GLM filled the bowl with glowing liquid that actually sloshes — the most convincing 'liquid in a bowl'. Opus's particles glowed but clumped to the centre. Kimi's collapsed into a tiny blob.
The winner on Fluid
GLM-5.2 took gold on this task. winner · best liquid.
What I saw: GLM filled the bowl with glowing liquid that actually sloshes — the most convincing 'liquid in a bowl'. Opus's particles glowed but clumped to the centre. Kimi's collapsed into a tiny blob.
See GLM-5.2's full model card: /models/glm.
Every attempt — live, playable
Side by side. Click any tile to run that model's actual one-shot HTML in a new tab.
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVEHow I scored Fluid — methodology
Three axes, 0–10 each, averaged. Runs: drop the .html in a browser; if it opens to a broken page, it scores zero. Hits the brief: did the model ship the thing the prompt asked for, or a different thing it found easier. Looks good: visual polish, motion, interactivity — where most of the gap between gold and silver lives.
My scores trace back to the source comparison guides on agentos.guide. See the full methodology page for data provenance, including which source guide each cell's score came from.
Related
More sim benchmarks: all tasks in the Sim category · See the best AI model for Fluid · Back to the leaderboard
Run this stack yourself.
Every demo on this bench was built inside the Agent Operating System — one prompt, one shot, single HTML file out. The Agent OS, the prompts, the templates, the weekly walkthroughs and 4,000+ founders shipping with it every day all live inside the AI Profit Boardroom.