Benchmarks
Top AI models on BazaarLink by Artificial Analysis Intelligence Index
X-axis: input price in NT$ / 1M tokens · top-left = smarter and cheaper
Full leaderboard
Top 10 models callable on BazaarLink · priced in TWD · invoice available
FAQ
How is the Artificial Analysis Intelligence Index defined?
It is a single 0-100 score set by the independent third party Artificial Analysis, combining public benchmarks such as MMLU-Pro, GPQA Diamond, Humanity's Last Exam, AIME, LiveCodeBench and SciCode. A higher number means stronger overall capability. Scores are taken directly from AA's public data and updated daily.
How is output speed (tok/s) defined?
tok/s is the number of tokens a model produces per second, the median output throughput measured by Artificial Analysis. Higher is faster. This is generation speed, distinct from time-to-first-token (TTFT).
How do I use these models in Taiwan, and what do they cost?
Call them over BazaarLink's OpenAI-compatible API. Pricing is shown in TWD and invoices are available for business purchases. Actual unit prices follow the table above and update daily with the exchange rate.
Does this ranking reflect Traditional Chinese ability?
Not directly. The Artificial Analysis Intelligence Index is a global, mostly English benchmark measuring general reasoning, math and coding, and does not test Traditional Chinese comprehension or Taiwan-local knowledge. For Traditional-Chinese / Taiwan context, see the Ministry of Digital Affairs' AI Evaluation Center (AIEC, aiec.org.tw), which evaluates models on Taiwan values and high-school exam Chinese and social-studies benchmarks.
Intelligence and speed data from Artificial Analysis. TWD prices are BazaarLink's actual selling prices, updated daily.（Artificial Analysis・1 USD = 32.08 TWD）