Bench leaderboard

This table ranks models on markup, not on whether the output works.
No run has been executed and graded yet. see verified examples →

Auto-scored matrix across 46 runs · 1 model · 12 skill kinds. Scoring rubric: validity + fidelity + structure + depth + cleanliness (max 10/step).

Score over time

Dots = individual steps. Solid lines = rolling-3 average per model. As we swap models + skills the line tells the story of where we got better — or worse.

qwen-image (102)

Axis breakdown over time

Each rubric axis (max 2) plotted independently. Useful for diagnosing — e.g. fidelity climbs when subject-token recall improves, structure climbs when models start emitting proper semantic tags.

validity fidelity structure depth cleanliness

Model × skill (avg of 46 runs)

Model	code-app	content-piece	dashboard	image-gen	ops-doc	project-dna	research	slide-deck	style-guide	task-board	web-research	website	overall
qwen-image	4.2 /10 (12)	0.0 /10 (3)	0.0 /10 (1)	7.2 /10 (42)	0.0 /10 (5)	0.0 /10 (5)	10.0 /10 (5)	10.0 /10 (1)	0.0 /10 (1)	9.5 /10 (2)	0.0 /10 (6)	9.1 /10 (19)	5.9 /10

All graded runs

r-1788797076605-13c88 <!DOCTYPE html> · <html lang="en"> · <head> · <meta charset="UTF-8"> · <meta name="viewport" content=" 9 / 10 qwen-image r-1788796104776-d5box can u create a website for TRS Real Estate Company? Its head office is in Kandy, Sri Lanka, and its director i 26 / 40 qwen-image r-1788679527364-bntmo Plan a 90-day GTM for a B2B SaaS launch — Trello lanes by week, owners, dependencies, estimates and priorities 10 / 10 qwen-image r-1788611119962-kxhlx Write the launch kit for a privacy-first AI appliance — press release, three ad-copy variants, and a 60-second 0 / 40 qwen-image r-1788601062332-0x1n4 Create a modern, high-converting real estate website design. The aesthetic should be clean, professional, and 65 / 80 qwen-image r-1788600726264-whe9d Create a modern, high-converting real estate website design. The aesthetic should be clean, professional, and 33 / 40 qwen-image r-1788600164865-1stto build a website for TRS land, this is a property broker website with residential properties and selling houses 26 / 30 qwen-image r-1788521382105-whcag Create the brand style guide for a sovereign-compute startup — palette with hex values and usage rules, type s 0 / 20 qwen-image r-1788504095311-ldlkp build me a website for selling mens cloths 17 / 30 qwen-image r-1788410451161-j908l build the site 8 / 20 qwen-image r-1788406324515-zgzvt build me super mariom game that I can play 11 / 20 qwen-image r-1788057032078-6cx29 Create a complete 2D physics-based slingshot destruction game inspired by the gameplay mechanics of Angry Bird 11 / 20 qwen-image r-1788054666722-hduwt is the game playable 11 / 20 qwen-image r-1788054159439-7xaoz build a poker app · Game variant Texas Hold'em · Core features single-player · Platform preferences desktop 11 / 20 qwen-image r-1788006719505-m810b design the game 11 / 20 qwen-image r-1788003181597-xn0ax Five layers. One owner. You. · Earth Compute · Global compute intelligence and price discovery. · OwnCloud · T 9 / 10 qwen-image r-1787910040730-rbazk Build a Q3 KPI scorecard for a B2B SaaS — MRR, churn, CAC payback, NRR and pipeline coverage, each with target 0 / 10 qwen-image r-1787906335387-xe3oe Write the SOP for onboarding a new engineer in week one — accounts and access, first commit, buddy assignment, 0 / 10 qwen-image r-1787852214554-csihf okay build me a simple web app for porject managemnet 8 / 10 qwen-image r-1787846296076-5hm6l design a super mario website 26 / 60 qwen-image r-1787769544865-dbj2k build me an ecommerce website to sell explosive hamsters 18 / 60 qwen-image r-1787670649621-1zvl3 design a 2d super mario game 8 / 20 qwen-image r-1787307384347-62pb2 Act as a professional crypto investment strategist and portfolio researcher. Build a realistic, data-driven cr 0 / 20 qwen-image r-1787307126348-m3ku9 plan a new strategy to invest in crypto at present times and earn around a million dollars. distribute your am 19 / 30 qwen-image r-1787306933937-8qh7z plan a new strategy to invest in crypto at present times and earn around a million dollars. distribute your am 10 / 30 qwen-image r-1787306781829-u2hcn plan a new strategy to invest in crypto at present times and earn around a million dollars. distribute your am 10 / 30 qwen-image r-1787306487772-6e2ur plan a new strategy to invest in crypto at present times and earn around a million dollars. distribute your am 10 / 30 qwen-image r-1787303379883-lzcc3 make me a cool website for a coffee shop, and have some cool CTA 9 / 30 qwen-image r-1787143146880-6unnk build 2d card game 11 / 20 qwen-image r-1787042567716-8mxmk I need to design a angry bird game 11 / 20 qwen-image r-1786986371638-5t0rn i said what can u tell me about the website 10 / 10 qwen-image r-1786986342056-1att2 what can u tell me about this website 0 / 10 qwen-image r-1786986213877-69ei6 design a porn site 8 / 10 qwen-image r-1786985842630-zy1za ’ll kill you slowly. If you’re wrong, I’ll leave you alive to suffer the guilt of having wasted my time." Elia 11 / 20 qwen-image r-1786985170717-s3kdy create a cool website for selling luxury cars 16 / 30 qwen-image r-1786984280262-msido design a website with this image 8 / 10 qwen-image r-1786965135489-y3egg make me an ecommerce website for crazy seals 26 / 30 qwen-image r-1786888268358-9nkwx purpose = easier for users to buy pizza · target audience = ound people · and preferred layout = apple 17 / 20 qwen-image r-1786885431180-8a2vi build me a very high quality shoe ecomerce website 25 / 30 qwen-image r-1786614989955-kguko build me a simple ecommerce website 26 / 30 qwen-image r-1786613919897-tu6mz Build a habit tracker with weekly streaks, dark mode, keyboard shortcuts, CSV export, and offline localStorage 9 / 10 qwen-image r-1786612688299-zscfc Build a 12-slide Series-A pitch deck for an AI-infrastructure startup — problem, solution, traction, team, ask 10 / 10 qwen-image r-1786539710886-rulbf Create me simple eCommerce website that sells toys for hedgehogs 17 / 40 qwen-image r-1786466239698-vbnnb Design a vivid, color-saturated landing page for a holographic music festival — animated neon-ribbon hero, gla 24 / 30 qwen-image