Get the Agent OS + join 4,000+ founders inside the AI Profit Boardroom → Join AIPB ($59/mo)
Meta

Muse Spark 1.2

Meta's coding reasoning model — co-trained with its own agent, 1M-token window.

Context1,000,000 tokens
Pricing$1.25 in / $4.25 out per 1M
Tasks tested50
Avg score7.47/10 average
Medals🥇0 🥈3 🥉4
Release2026-08-05
Official vendor source
Muse Spark 1.2 is built by Meta — see the vendor's own product page, pricing, and docs at developer.meta.com/ai/models/muse-spark.
Visit developer.meta.com/ai/models/muse-spark →

What is Muse Spark 1.2?

Muse Spark 1.2 is the Meta frontier model with a 1,000,000 tokens context window, released 2026-08-05. Tagline: Meta's coding reasoning model — co-trained with its own agent, 1M-token window.. Official source: developer.meta.com/ai/models/muse-spark.

Pricing detail. Meta's coding-optimized reasoning model, released 2026-08-05 beside the Muse Code agent. $0.15/1M cached input. Contributor tier is token-rate-limited in a rolling 5-hour window. Benched release-day via OpenRouter (meta/muse-spark-1.2, first-party listing); Opus 4.8 judged every real rendered poster, same rubric as the whole field.

How I use it inside the Agent OS. Cloud coder via OpenRouter; the Muse Code agent (one-command install) is its native harness.

What I built with Muse Spark 1.2

Every model on Goldie Bench gets the same fixed prompt set — one shot, single HTML file out — and I score the result 0–10 inside the Agent Operating System. Here's what Muse Spark 1.2 shipped on the bench: 50 one-shot demos across 1,000,000 tokens of context. Of those, 50 are scored against the field with my honest 0–10 from the source guides at agentos.guide.

Strengths

  • Generative art & shader-feel scenes (fractal 8.7, aurora/galaxy/matrix/synthwave 8.6)
  • Full app chrome one-shot (macOS-clone desktop 8.6)
  • Fast one-shots — most builds landed in 45-80s
  • 1M context for whole-repo work

Trade-offs

  • 3D game worlds often render black/empty (dragonrealm 2.5, dogfight 3.0, doom 3.5)
  • Open-world briefs collapse to HUD-only shells
  • Reasoning tokens billed as output

Best for

  • Generative-art visuals
  • Dashboard & app-shell one-shots
  • Long-context refactors (1M window)

Every demo by Muse Spark 1.2

50 live demos, sorted by category. Click any tile to play the actual one-shot result. Verdicts and 0–10 scores are pulled from the source guides where I posted them publicly.

Muse Spark 1.2 one-shot build of Crypt — GoldieBench AI benchmark screenshot▶ LIVE
Crypt
Game
The HUD is polished and on-brief (torch fuel, vitalis, kills, coffin altar), but the rendered scene looks flat and bright rather than torch-lit and dark — lighting is washed-out with pale beige walls, floating spheres, and no atmospheric torch glow, badly undermining the dungeon-crawler mood.
Muse Spark 1.2 one-shot build of Dragonrealm — GoldieBench AI benchmark screenshot▶ LIVE
Dragonrealm
Game
Strong, polished frozen-world render with cohesive HUD (health/stamina/compass/kills), a nicely stylized low-poly hero, snow terrain, pines and rocks with clean shadows — clearly on-brief and shippable, though it reads more diorama than expansive Skyrim vista and no dragon/enemies visible in the shot keeps it just short of the top.
Muse Spark 1.2 one-shot build of Doom — GoldieBench AI benchmark screenshot▶ LIVE
Doom
Game
HUD is polished and DOOM-themed (health, kills, automap with dot markers, shotgun ammo/shells), but the actual raycast view is a flat empty brown gradient with no visible walls or monsters in front of the player — the core 'monsters chasing you in a maze' gameplay isn't demonstrated on screen. The minimap shows enemies exist, but the first-person scene is essentially blank aside from the weapon.
Muse Spark 1.2 one-shot build of Racing — GoldieBench AI benchmark screenshot▶ LIVE
Racing
Game
Strong: crisp third-person 3D craft on a curved track with lane markers, obstacles (crates/hexagon), trees, minimap, and a cohesive premium HUD with health/speed/boost meters. Weak: camera framing is partly obscured by a large white wall on the right and the craft looks slightly off-track, slightly hurting immersion.
Muse Spark 1.2 one-shot build of Skyrim — GoldieBench AI benchmark screenshot▶ LIVE
Skyrim
Game
The HUD is polished and thematically on-brief (compass, health/stamina/magicka bars, Cinzel fonts), but the 3D world is bare and unconvincing—flat washed-out terrain, only a couple of low-poly trees and rocks, and a giant floating gray monolith that looks like a broken mesh. No visible enemies, buildings, or sense of an explorable Skyrim world, making it feel unfinished versus the field's best.
Muse Spark 1.2 one-shot build of Twilightvale — GoldieBench AI benchmark screenshot▶ LIVE
Twilightvale
Game
Renders a clean 3D world with cottage, trees, character, terrain, minimap and full RPG HUD (health/stamina/exp, quest, kills/gold) — polished and cohesive purple-gold dusk palette. Weaker on the 'weather/terrain' ambition (looks snowy but flat lighting, floating white cubes read as placeholder props), and combat isn't visibly demonstrated in the shot.
Muse Spark 1.2 one-shot build of Voxelcraft — GoldieBench AI benchmark screenshot▶ LIVE
Voxelcraft
Game
Renders a functional voxel world with block highlight, water, polished HUD (hearts, clock, hotbar with 9 block types) and clear controls; but the dark sky reads as broken/black at 05:23 AM 'DAY', the terrain looks blocky-flat with visible clipping, and it lacks the visual polish/texturing of top entries.
Muse Spark 1.2 one-shot build of Gtafoot — GoldieBench AI benchmark screenshot▶ LIVE
Gtafoot
Game
Renders cleanly with a polished HUD (health/ammo/kills, wanted stars, functional minimap, cash/cops) and an on-brief third-person city scene with pedestrians, hydrant cover props, fountain plaza, and lampposts. Weak spots: the foreground pink car/box is oversized and awkwardly framed near the crosshair, buildings look flat and sparse, and the overall look reads generic-blocky rather than a standout GTA feel.
Muse Spark 1.2 one-shot build of Gtadrive — GoldieBench AI benchmark screenshot▶ LIVE
Gtadrive
Game
Clean 3D render with a stylized city, drivable car, working HUD (health/wanted/stars/speed), functional minimap with cop dots, and full control legend — polished and clearly on-brief. Slightly generic driving feel and sparse traffic visible in-frame keep it just below the very top.
Muse Spark 1.2 one-shot build of Aipbpromo — GoldieBench AI benchmark screenshot▶ LIVE
Aipbpromo
Page
Strong cinematic hero scene with polished glassmorphic stage, animated glow orbs, video-player chrome (timeline, chapters, transport controls) and clear on-brand copy — very shippable. Minor flaws: the time readout shows a broken '-1:-1 / 00:32' and the top-right hint overlaps into the header, keeping it just below task-winner level.
Muse Spark 1.2 one-shot build of Parachute — GoldieBench AI benchmark screenshot▶ LIVE
Parachute
Game
Renders cleanly with a polished HUD (altitude, vertical speed, distance-to-clearing, minimap) and a nicely deployed canopy over a vast forested landscape — on-brief for the canopy-steering phase. Slightly generic terrain and the tiny distant clearing hurt visual punch, and the atmospheric depth is thinner than the field's best.
Muse Spark 1.2 one-shot build of Flightsim — GoldieBench AI benchmark screenshot▶ LIVE
Flightsim 🥉
Game
Strong render: clean 3D plane, textured runway with markings, terrain with trees, and a genuinely rich flight HUD (airspeed/altitude/VS tape, heading compass, throttle/hull, artificial horizon, gear/kills chips). Slightly generic terrain and low-detail plane keep it just below the field's best, but it's clearly shippable and on-brief.
Muse Spark 1.2 one-shot build of Arcade — GoldieBench AI benchmark screenshot▶ LIVE
Arcade
Game
Renders a polished 3D arena with strong HUD (radar, meters, score) and clean neon styling, but it drifts off-brief — the task asked for classic tetris/breakout/snake and this is a 3D twin-stick 'grid breaker', plus the scene reads as sparse and unclear with floating cones and thin lines rather than a recognizable arcade game.
Muse Spark 1.2 one-shot build of Dogfight — GoldieBench AI benchmark screenshot▶ LIVE
Dogfight
Game
Strong, polished chase-cam dogfight: crisp HUD with meters/radar/lock indicator, clean 3D plane and volumetric clouds, visible bandits and a lock-on target — very shippable. Slightly held back from top by generic sky-only environment and no visible combat action (tracers/effects) in the frame.
Muse Spark 1.2 one-shot build of Dragonflight — GoldieBench AI benchmark screenshot▶ LIVE
Dragonflight
Game
HUD is genuinely polished — health/fury/speed meters, altitude/rings/combo, score+best, minimap, and control hints all look shippable — but the actual 3D scene is broken: the camera is buried inside a giant cone/horn with no visible dragon body, no neon rings, no canyon, and no sky. The core gameplay view fails to render on-brief.
Muse Spark 1.2 one-shot build of Game — GoldieBench AI benchmark screenshot▶ LIVE
Game
Game
Strong: fully rendered Three.js 3D arena shooter with a cohesive premium sci-fi HUD (hull, thruster, score, hostiles), crosshair, and enemy ships/rovers on a lit terrain. Weak: the bright washed-out palette makes low-poly rocks and hero read a bit flat/plain, keeping it just shy of the top tier.
Muse Spark 1.2 one-shot build of Neonblaster — GoldieBench AI benchmark screenshot▶ LIVE
Neonblaster
Game
Polished HUD with hull/shield/overheat meters, minimap, and crisp control legend renders cleanly, and the 3D low-poly asteroid field with a proper player ship and shadows looks juicy and on-brief. Loses the top slot on visual clarity—the enemies read as generic gray rocks rather than menacing foes and there's no visible boss or heavy neon/particle punch in this frame to match the best in field.
Muse Spark 1.2 one-shot build of Neoncity — GoldieBench AI benchmark screenshot▶ LIVE
Neoncity
Game
Renders a clean, cohesive cyberpunk drive scene with a solid vehicle, neon-lit road markings, buildings, working minimap and a polished HUD (integrity, threats, credits, controls). Strong and shippable, but the city feels a bit sparse/dark on neon signage and the mid-ground buildings are plain, keeping it just short of the top-tier best.
Muse Spark 1.2 one-shot build of Neonracer — GoldieBench AI benchmark screenshot▶ LIVE
Neonracer
Game
Strong render: 3D neon craft on a glowing cyan/magenta-lined road with atmospheric fog, city silhouettes, colored point-light bloom, and a polished HUD (integrity/velocity/boost meters, minimap, controls strip). Weaknesses vs the field's best — vapor-trail particles aren't clearly visible in this frame and the scene reads slightly static, keeping it just under top-tier.
Muse Spark 1.2 one-shot build of Nordiccrypt — GoldieBench AI benchmark screenshot▶ LIVE
Nordiccrypt
Game
Strong, polished HUD (vitality/breath, minimap, crypt score, control hints) and the scene renders with pillars, draugr, crate/rock props and a glowing relic pyramid — but the lighting reads bright and daylit rather than torch-lit Nordic dungeon gloom, undercutting the core 'torch-lit crypt' brief and leaving it generic.
Muse Spark 1.2 one-shot build of Outrun — GoldieBench AI benchmark screenshot▶ LIVE
Outrun
Game
The HUD renders cleanly with polished synthwave styling, but the entire 3D scene — road, sun, mountains, car — is a black void, so the actual outrun driving game never appears. A functional shell around a non-rendering canvas fails the core brief.
Muse Spark 1.2 one-shot build of Pool — GoldieBench AI benchmark screenshot▶ LIVE
Pool
Game
Renders a clean 3D pool table with properly racked colorful balls, cue ball, pockets, radar minimap, and a polished HUD — clearly on-brief and playable. Weakened by an odd flat teal background/environment, an off-kilter cue prop, and gimmicky 'HP/ENEMIES/haunted' framing that adds clutter rather than billiards depth.
Muse Spark 1.2 one-shot build of Raycaster — GoldieBench AI benchmark screenshot▶ LIVE
Raycaster
Game
Strong, polished HUD/menu with rich sci-fi framing, minimap, weapon HUD, and a rendering 3D scene visible behind the blurred start overlay — but the screenshot only shows the pre-game menu, so the actual maze walkthrough and enemy/exit gameplay aren't verifiable on-screen, holding it below the field's best.
Muse Spark 1.2 one-shot build of Rpg — GoldieBench AI benchmark screenshot▶ LIVE
Rpg
Game
Strong, polished 3D top-down world with lovely low-poly trees/houses, working minimap, HP/gold/kills HUD, inventory hotbar and combat feedback ('SWING!') — genuinely shippable. Weak spots: player and enemies read as tiny generic blobs rather than sprites, and the world feels a touch sparse/empty in the mid-field, keeping it just below the top.
Muse Spark 1.2 one-shot build of Matrixrain — GoldieBench AI benchmark screenshot▶ LIVE
Matrixrain
Other
Strong 3D take on matrix rain with real depth, parallax columns, glowing floor grid and a polished glassmorphic control panel — head glyphs and trails read well; slightly weakened by the plain green monolith obscuring center streams and a somewhat overpowering floor glow, but overall a shippable, near-best entry.
Muse Spark 1.2 one-shot build of Mlx Speedtest — GoldieBench AI benchmark screenshot▶ LIVE
Mlx Speedtest
Other
Strong polished dashboard shell with clean panels, live metrics, sparklines and a benchmarking task list that reads convincingly on-brief. Weakened by the empty 3D center canvas (just faint ring, no visible model/animation) and blank throughput timeline chart, plus raw unformatted tok/s numbers, which keep it below the field's best.
Muse Spark 1.2 one-shot build of Neonsnake — GoldieBench AI benchmark screenshot▶ LIVE
Neonsnake
Other
The 3D perspective grid and neon 'LINK SEVERED' modal render with polish, but the game state shows immediate game-over (Score 0, Length 3, Tasks 0/12) with no snake visible — it evidently auto-ran into a wall on load, and most HUD text is barely legible/washed out. Strong aesthetic direction undercut by a broken-looking initial state.
Muse Spark 1.2 one-shot build of Landing — GoldieBench AI benchmark screenshot▶ LIVE
Landing
Page
Polished dark SaaS landing with strong typography, tasteful glass cards, gradient testimonial block and a 3D torus knot that adds depth; but the WebGL object overlaps and obscures the middle feature card's text, and the hero appears scrolled past so the headline reads mid-section rather than a clean top-of-page hook.
Muse Spark 1.2 one-shot build of Webos — GoldieBench AI benchmark screenshot▶ LIVE
Webos
Page
Strong, cohesive macOS-style build with all three apps functional and visible (Notes with toolbar, Paint with working strokes/palette, Terminal responding to ls/echo), plus a Files app, glossy dock, topbar with live clock, and a gorgeous animated gradient wallpaper. Only minor nit is some window overlap clutter, but the polish and completeness clearly match the field's best.
Muse Spark 1.2 one-shot build of Blackhole — GoldieBench AI benchmark screenshot▶ LIVE
Blackhole
Sim
Strong Interstellar-style accretion disk with warm temperature gradient, event horizon shadow and photon ring reads clearly, and the HUD is polished; but the lensing overreaches — the massive glowing arc across the bottom looks like an artifact rather than convincing lensed sky, undercutting the physics claim.
Muse Spark 1.2 one-shot build of Boids — GoldieBench AI benchmark screenshot▶ LIVE
Boids
Sim
Strong 3D boids with visible flock clustering, distinct bird meshes, a red hawk predator, bounding box, grid ground and a polished control/telemetry UI; slightly held back by 32 FPS and somewhat loose clustering rather than tight emergent flocks.
Muse Spark 1.2 one-shot build of Cloth — GoldieBench AI benchmark screenshot▶ LIVE
Cloth
Sim
Strong, convincing verlet drape with natural folds and soft shadow over the sphere, polished UI with multiple object presets and toggles; weakened slightly by the 20fps counter suggesting perf strain and a somewhat plain floor/lighting compared to the field's best.
Muse Spark 1.2 one-shot build of Fluid — GoldieBench AI benchmark screenshot▶ LIVE
Fluid
Sim
Gorgeous, richly-colored swirling particle vortex with strong bloom, glowing droplets, and polished glass UI that reads exactly on-brief; only concern is the 9 FPS readout suggesting real performance may lag despite the stunning still.
Muse Spark 1.2 one-shot build of Fractal — GoldieBench AI benchmark screenshot▶ LIVE
Fractal 🥈
Sim
Strong: a beautifully rendered GPU Julia set with smooth coloring, orbit-trap glow, and a polished glassmorphic control panel (mode toggle, iterations, palettes, live zoom/center stats). Weak: nothing major visible — a very complete, on-brief, task-winning build.
Muse Spark 1.2 one-shot build of Galaxy — GoldieBench AI benchmark screenshot▶ LIVE
Galaxy 🥉
Sim
Strong render — clearly defined spiral arms with a bright glowing core, a warm-to-cool color gradient disk, and polished glassy UI (title, color mode, arm slider, hint bar). The composition is genuinely galaxy-like and on-brief with rich interactivity (swirl/zoom/warp), making it a task-topping build.
Muse Spark 1.2 one-shot build of Orbit — GoldieBench AI benchmark screenshot▶ LIVE
Orbit 🥉
Sim
Strong on-brief render: glowing 3D bodies with size/color mass tiers, orbital rings, starfield, and a comprehensive control panel (presets, G/softening/timewarp, trails/vectors, energy readout). Polished and shippable; only mild concern is whether the physics/trails feel truly dynamic, but the visual and UX quality tops the field.
Muse Spark 1.2 one-shot build of Particleforge — GoldieBench AI benchmark screenshot▶ LIVE
Particleforge 🥉
Sim
Strong render — a glowing swirling galaxy of thousands of particles over a subtle planet horizon, with a polished, cohesive UI (presets, gravity brush modes, live stats, cursor well). Particles read as blocky squares rather than soft glowing points, keeping it just shy of the task best.
Muse Spark 1.2 one-shot build of Pathtracer — GoldieBench AI benchmark screenshot▶ LIVE
Pathtracer
Sim
Genuinely rendering a noisy-but-converging Cornell Box with recognizable red/green walls, area light, and metal/glass spheres — the Monte-Carlo grain and progressive accumulation read as authentic path tracing, backed by a polished UI with presets, live SPP/rays stats and a light-path viz. Slightly weak on visible caustics/refraction fidelity and the geometry looks a touch muddy, keeping it just shy of the top.
Muse Spark 1.2 one-shot build of Reactiondiff — GoldieBench AI benchmark screenshot▶ LIVE
Reactiondiff
Sim
Strong, polished UI with full preset/palette/slider controls and a genuinely evolving Gray-Scott sim visible in the render; the pattern looks a bit soft/blobby rather than crisp Turing structure and coverage is sparse, keeping it just short of the field's best.
Muse Spark 1.2 one-shot build of Solar — GoldieBench AI benchmark screenshot▶ LIVE
Solar
Sim
Strong render — glowing sun with layered corona, elliptical orbits with visible eccentricity, a dense asteroid belt, and a rich, well-organized UI (time/distance scale, focus, legend, live planet stats). Slightly generic planet textures and some crowding hold it just under the top tier, but it's clearly shippable and on-brief.
Muse Spark 1.2 one-shot build of Wormhole — GoldieBench AI benchmark screenshot▶ LIVE
Wormhole
Sim
Strong polished HUD/UI with velocity, warp drive and hint pills, and a real tube-shader tunnel is present, but the render is badly overblown — huge white bloom washes out the tunnel depth and edges bleed into hard pink/cyan blocks, killing the sense of a clean flythrough. Concept and controls are there but the visual execution reads noisy and clipped rather than the deep, coherent wormhole the top builds achieve.
Muse Spark 1.2 one-shot build of Aurora — GoldieBench AI benchmark screenshot▶ LIVE
Aurora 🥈
Visual
Strong 3D scene with vivid, silky aurora curtains over snowy dunes, silhouetted pine trees, moon and a frozen lake, plus a polished glassy HUD with Kp badge and control chips. Rich composition and glowing shader ray detail push it past the field's best; only minor nit is the somewhat flat foreground snow expanse.
Muse Spark 1.2 one-shot build of Fireworks — GoldieBench AI benchmark screenshot▶ LIVE
Fireworks
Visual
Strong 3D cityscape scene with polished UI, gradient title, and orbit controls, but the screenshot shows rockets rising rather than actual exploding bursts, making the signature fireworks moment absent and the scene feel a bit static/washed-out.
Muse Spark 1.2 one-shot build of Lavalamp — GoldieBench AI benchmark screenshot▶ LIVE
Lavalamp
Visual
Renders a polished 3D scene with nice metal cap/base, glowing coil, and organic blobs, but the glass cylinder is nearly invisible so the blobs float in a void and the color reads pale cream rather than fiery lava, breaking the illusion. Strong lighting and UI polish, but the missing glass container and off-brief coloring keep it below the top tier.
Muse Spark 1.2 one-shot build of Matrix — GoldieBench AI benchmark screenshot▶ LIVE
Matrix 🥈
Visual
Gorgeous dense rain with bright white heads, proper katakana glyphs, and a cohesive cyberpunk HUD (stream status, carrier signal, control bar) that elevates it well past a generic canvas demo. Strong glow, vignette, and interactivity hooks make it a task winner; the only minor nit is the 3D Three.js layer being mostly hidden behind the dominant rain.
Muse Spark 1.2 one-shot build of Plasma — GoldieBench AI benchmark screenshot▶ LIVE
Plasma
Visual
Renders a vivid, flowing plasma field with a polished glass UI, five-swatch palette morpher, and clear ripple/randomize controls that fully cover the brief. Strong composition and color, but the plasma reads more as smooth gradient blobs than the classic hypnotic interference pattern, keeping it just shy of task-winner.
Muse Spark 1.2 one-shot build of Synthwave — GoldieBench AI benchmark screenshot▶ LIVE
Synthwave
Visual
Nails the brief with a gorgeous gradient sky, retro-lined sun, layered mountains, glowing cyan grid and polished neon typography — a complete, on-genre scene. Only minor weakness is the awkward palm silhouettes reading as odd sticks, but overall it tops the field.
Muse Spark 1.2 one-shot build of Terrain — GoldieBench AI benchmark screenshot▶ LIVE
Terrain
Visual
UI is polished and well-designed (Flight Deck, biome/ruggedness controls, live stats, compass), but the rendered terrain is nearly washed-out white with only faint distant peaks and a huge blank foreground, so the actual 3D terrain reads as flat and underwhelming rather than a compelling explorer.
Muse Spark 1.2 one-shot build of Voxel — GoldieBench AI benchmark screenshot▶ LIVE
Voxel
Visual
Renders a recognizable voxel landscape with trees, water, and a polished UI, but the washed-out overexposed lighting flattens the blocks and makes water/edges read as blank white patches, hurting the crisp Minecraft look. Solid structure and clean HUD, but the visual clarity is well below the field's best.
Muse Spark 1.2 one-shot build of Waves — GoldieBench AI benchmark screenshot▶ LIVE
Waves
Visual
The polished UI (control panel, buoy, islands, ripples) is impressive and clearly on-brief, but the actual ocean surface looks nearly flat and washed-out white with no visible wave undulation, and the 10 FPS readout signals performance trouble — the core 'animated wave simulation' isn't convincingly delivered in the render.
every demo, in a grid · click any one to play

Compare Muse Spark 1.2 against every other model

Every head-to-head featuring Muse Spark 1.2. Verdicts shown for scored pairs.

Muse Spark 1.2 vs Fusion
Fusion leads 39–8
Muse Spark 1.2 vs Claude Opus 5
Claude Opus 5 leads 36–7
Muse Spark 1.2 vs Hermes MoA
Hermes MoA leads 29–14
Muse Spark 1.2 vs GPT-5.6 Sol
GPT-5.6 Sol leads 34–7
Muse Spark 1.2 vs Claude Fable 5
Claude Fable 5 leads 22–21
Muse Spark 1.2 vs Qwen 3.8
Qwen 3.8 leads 29–6
Muse Spark 1.2 vs Grok
Grok leads 23–20
Muse Spark 1.2 vs MiniMax M3
MiniMax M3 leads 24–23
Muse Spark 1.2 vs Fugu Ultra
Fugu Ultra leads 23–19
Muse Spark 1.2 vs Kimi K3
Kimi K3 leads 32–13
Muse Spark 1.2 vs GLM-5.2
Muse Spark 1.2 leads 26–21
Muse Spark 1.2 vs Fugu Mini
Muse Spark 1.2 leads 22–14
Muse Spark 1.2 vs Opus 4.8
Muse Spark 1.2 leads 28–19
Muse Spark 1.2 vs Kimi K2.7
Muse Spark 1.2 leads 18–7
Muse Spark 1.2 vs Qwable 5 27B Coder
Muse Spark 1.2 leads 27–13
Muse Spark 1.2 vs Gemini 3.6 Flash
Muse Spark 1.2 leads 27–16
Muse Spark 1.2 vs Claude Sonnet 5
Muse Spark 1.2 leads 25–18
Muse Spark 1.2 vs Qwen 3.7
Muse Spark 1.2 leads 31–16
Muse Spark 1.2 vs Fugu Ultra 1.1
Tied 11–11
Muse Spark 1.2 vs Inkling
Muse Spark 1.2 leads 41–9
Muse Spark 1.2 vs Agents-A1
Muse Spark 1.2 leads 37–8
Muse Spark 1.2 vs Gemma 4 12B · MLX
Muse Spark 1.2 leads 36–5
Muse Spark 1.2 vs Laguna XS 2.1
Muse Spark 1.2 leads 39–2
Muse Spark 1.2 vs Qwythos 9B
Muse Spark 1.2 leads 39–3
Muse Spark 1.2 vs LongCat-2.0
LongCat-2.0 leads 4–0
Muse Spark 1.2 vs Hy3
Muse Spark 1.2 leads 7–0
Muse Spark 1.2 vs Gemma-4 12B Coder
Muse Spark 1.2 leads 6–0
Muse Spark 1.2 vs DeepSeek V4 Flash
50 shared tasks · unscored
Muse Spark 1.2 vs DeepSeek V4 Pro
50 shared tasks · unscored
Muse Spark 1.2 vs Kimi K2.7 · Fast
47 shared tasks · unscored
Muse Spark 1.2 vs Kimi K2.7 · No-Think
47 shared tasks · unscored
Muse Spark 1.2 vs Kimi K2.7 · Quality
47 shared tasks · unscored
Muse Spark 1.2 vs Ornith 1.0
42 shared tasks · unscored
Muse Spark 1.2 vs Claude Mythos 5
Reference-only
Muse Spark 1.2 vs Kilo Code
Reference-only

See all 66 comparisons across every model →

Quick pill index

Direct comparisons against every other scored model on the bench:

Muse Spark 1.2 vs Fusion Muse Spark 1.2 vs Claude Opus 5 Muse Spark 1.2 vs Hermes MoA Muse Spark 1.2 vs GPT-5.6 Sol Muse Spark 1.2 vs Claude Fable 5 Muse Spark 1.2 vs Qwen 3.8 Muse Spark 1.2 vs Grok Muse Spark 1.2 vs MiniMax M3 Muse Spark 1.2 vs Fugu Ultra Muse Spark 1.2 vs Kimi K3 Muse Spark 1.2 vs GLM-5.2 Muse Spark 1.2 vs Fugu Mini Muse Spark 1.2 vs Opus 4.8 Muse Spark 1.2 vs Kimi K2.7 Muse Spark 1.2 vs Qwable 5 27B Coder Muse Spark 1.2 vs Gemini 3.6 Flash Muse Spark 1.2 vs Claude Sonnet 5 Muse Spark 1.2 vs Qwen 3.7 Muse Spark 1.2 vs Fugu Ultra 1.1 Muse Spark 1.2 vs Inkling Muse Spark 1.2 vs Agents-A1 Muse Spark 1.2 vs Gemma 4 12B · MLX Muse Spark 1.2 vs Laguna XS 2.1 Muse Spark 1.2 vs Qwythos 9B Muse Spark 1.2 vs LongCat-2.0 Muse Spark 1.2 vs Hy3 Muse Spark 1.2 vs Gemma-4 12B Coder

Read more on agentos.guide:

Muse Spark 1.2 — frequently asked

What is Muse Spark 1.2?

Muse Spark 1.2 is Meta's AI model — Meta's coding reasoning model — co-trained with its own agent, 1M-token window. It has a 1M tokens context window and was released 2026-08-05.

How good is Muse Spark 1.2 at coding and one-shot builds?

On the GoldieBench one-shot build benchmark it averages 7.47/10 across 50 scored tasks, with 0 gold, 3 silver and 4 bronze medals.

How much does Muse Spark 1.2 cost?

$1.25 in / $4.25 out per 1M. Meta's coding-optimized reasoning model, released 2026-08-05 beside the Muse Code agent. $0.15/1M cached input. Contributor tier is token-rate-limited in a rolling 5-hour window. Benched release-day via OpenRouter (meta/

Where can I see Muse Spark 1.2 demos?

Every one-shot build is live and playable on this page and on the GoldieBench compare matrix — same prompt as every other model, no retries.

The same stack Julian uses

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.

4,000+founders
258documented wins
38countries
$59/momonthly