TASK 14 / HTML · V1 · Website

Bank Website

银行网站

4 runs · 14.A.01 · HTML

Compare works →
TASK 14 · 14.A.01 · HTMLBank Website
Open independent copy ↗
Bank Website

Preview note: Embedded storage and downloads can be restricted by browser sandboxing; use the standalone view for full behavior.

ORIGINAL PROMPT

Original prompt

PROMPT/Phase1_TEST_PROMPTS_BILINGUAL.md · The original prompt is kept verbatim

Source ↗
ORIGINAL PROMPT

Original prompt

Make me a website for a bank.
TRANSLATION

Chinese translation

为我制作一个银行网站。

Expected outcome

Because the original prompt is highly open-ended, a reasonable expected outcome is a credible, professional bank website that is clearly identified as a demonstration. It should include at least a branded home page, personal or business product overviews, core account, loan, or rate information, sign-in or account-opening entry points, and help and security guidance. The site should provide clear navigation, a trustworthy visual language, accurate information hierarchy, and responsive behavior. Any fictional rates, fees, authentication, or application flows must carry explicit disclaimers and must not imply real payments, insured deposits, or data transmission. Forms, calculators, and flows should function correctly on the front end, with attention to security messaging, privacy, accessibility, and financial formatting.

Additional verification guidance

  • Traverse primary navigation from the home page and verify links, menus, mobile drawers, breadcrumbs or current-page states, with no dead ends or broken links.
  • Review products, rates, fees, CTAs, sign-in/application, and help content for completeness and hierarchy; clearly disclose all fictional information and non-submitting forms.
  • Validate loan, savings, or other financial calculators using known examples and boundary values, checking currency, percentages, terms, rounding, and invalid input.
  • On desktop and mobile, inspect tables, forms, navigation, contrast, keyboard focus, labels, validation errors, and sensitive-data warnings; confirm no real submission occurs.
THREE INTERPRETATIONS

Same task, three model interpretations

All three results use the same width and the same preview ratio; no model is marked as best.

COMPARISON METRICS

Comparison metrics

Cross-model comparison must note differences in harness, budget and isolation.

MetricDeepSeek-V4.1-Flashrun-deepseek-v4-1-flash-exp-0910-task-14-r1Muse Spark 1.3run-muse-spark-1-3-task-14-r1GPT-5.6 Solrun-gpt-5-6-sol-task-14-r1K3run-k3-task-14-r1
Runtime23:042:009:3117:28
API calls12313627
Tool calls13712529
Failures / retries4001
Token throughput16.7M252.9K216.3K1.5M
AI review94/10088/10069/10091/100
RUN RECORDS

Run records

Two runs of the same task version can be compared side by side.

14.A.01 · run-deepseek-v4-1-flash-exp-0910-task-14-r1DeepSeek-V4.1-Flash● Archived
Runtime
23:04
Token throughput
16.7M
API calls
123
View archive record →
14.B.01 · run-muse-spark-1-3-task-14-r1Muse Spark 1.3● Archived
Runtime
2:00
Token throughput
252.9K
API calls
13
View archive record →
14.C.01 · run-gpt-5-6-sol-task-14-r1GPT-5.6 Sol● Archived
Runtime
9:31
Token throughput
216.3K
API calls
6
View archive record →
RELATED TASKS

Related tasks

View all tasks →