Plotly’s test puts Kimi K3.1 Pro near the top on data analysis, at a cost
Plotly tested Moonshot AI’s Kimi K3.1 Pro on its data-analysis exam, where it answered 42 of 43 questions correctly, up from 35 for Kimi K3. Plotly said the model also found all five hidden patterns and caught all three planted data-quality problems, but took a typical 2.5 minutes per question and cost €12.43 ($13.47) for the exam.
Bottom line — Kimi K3.1 Pro scored 42 of 43, but Plotly found it slower and costlier than some models with similar results.
Go deeper 6
-
Plotly said the score tied Grok 4.7 for third place among 34 models, behind GPT-6.1 Sol and GPT-6 Astra, which scored 43.
-
Plotly’s exam is a private benchmark using a synthetic wind-energy dataset; the company says it keeps the test private to protect its answer key.
-
Plotly reported that K3.1 Pro made about 73 tool calls on each hidden-pattern question, compared with about 18 for GPT-6.1 Sol.
-
Plotly said the exam cost about €0.29 ($0.31) per question at OpenRouter prices, with most of the bill coming from the model’s output.
-
Plotly measured about 16,500 output tokens per question and a typical response time of 2.5 minutes; it tested the standard model, not Moonshot’s faster Highspeed variant.
-
In a separate evaluation, Wallaby reported strong scores on specific Japanese and Turkish idiom tasks, but Turkish conversational implication scored 56.0% and German indirect-answer tasks 66.7%. Wallaby said these results apply to named datasets, not whole languages.