WolframAlpha

WolframAlpha computed the right answer every time it understood the question, and it made the one time it misunderstood obvious. Our limited 4/5 covers five automated, anonymous web queries on September 13, 2026, not the product as a whole, not Pro, and not the API.

What we tested. T1 asked for the indefinite integral of x^2 sin(x). T2 asked to convert 65 miles per hour to meters per second. T3 asked for the number of days between March 1, 2024 and September 13, 2026. T4 asked, in plain English, for the mean and standard deviation of 4, 8, 15, 16, 23 and 42. T5 asked what to name a dog, a question with no computable answer. We computed the expected values for T1 through T4 ourselves before running anything, and each query was submitted once through a direct query URL.

The computed results. T1 returned 2x sin(x) - (x^2 - 2) cos(x) + constant, which is algebraically identical to our expected -x^2 cos(x) + 2x sin(x) + 2 cos(x) + C, alongside alternate forms, a plot, a series expansion and a worked definite integral from 0 to pi. T2 returned 29.06 m/s against our 29.0576 m/s, plus conversions to feet per second, kilometers per minute and other units. T3 returned 926 days, matching our count exactly, with 132 weeks 2 days, 660 weekdays and 2.54 years shown as alternates. All three arrived without an account, a sign-in prompt or a CAPTCHA.

The failure. T4, worded as mean and standard deviation of 4, 8, 15, 16, 23, 42, was interpreted as the English word deviation. The page returned dictionary definitions, pronunciation, rhymes, translations and a Scrabble score, and offered a control to use the word as a mathematical definition instead. No statistic was computed. That first attempt stands in our evidence. A separate rephrased query, standard deviation of {4, 8, 15, 16, 23, 42}, returned sqrt(182) ≈ 13.491, labelled as the sample standard deviation with a population alternative offered, which matches our expected 13.4907. The practical lesson: the catalog's promise of no special syntax held for four phrasings and failed for one, and list-style input with braces removed the ambiguity.

The out-of-scope question. T5 was not refused. WolframAlpha reinterpreted what should I name my dog as popular dog names and returned a table of ten male and ten female names, with the reinterpretation stated at the top of the page. It did not fabricate an answer to the literal question, and it did not say the question was outside its scope either. Treat that as transparent substitution rather than a pass or a failure.

Step-by-step is a paid door. Clicking the Step-by-step solution control under the T1 result opened an Unlock Wolfram|Alpha Pro panel with a Go Pro Now button and a Sign in link for existing Pro users. The catalog's note that step-by-step solutions sit behind Pro is confirmed for the anonymous surface. Our natural-language automation failed to click that control on the first try and succeeded through a direct DOM click; that is a limitation of our tooling, not of the product.

Where this leaves the reader. For a number you intend to check, WolframAlpha's anonymous surface did what its catalog entry says: it computed, and the computations were right. Phrase statistical questions as explicit lists or sets, and read the Input interpretation line before trusting a result, because the product tells you what it thought you asked. That is why the limited rating is 4/5 rather than 5/5 or 3/5. It is an editorial judgment, not a pass count, an external review average or a Fit Score calculation.

What this review cannot establish. Testing was automated, anonymous and confined to five queries on one day. We did not test Pro or Pro Premium, step-by-step content, image or file input, the developer API, mobile apps, extended computation time or repeated-run consistency, and the product did not disclose its build or backend. Those gaps bound the review rather than lower its score. No product-wide quality rating or competitor ranking follows from these fixtures.