We asked the Consilium panel: what is the Gregorian date of 1 Ramadan 1448 AH per the Umm al-Qura calendar? Five frontier models answered independently — and produced three different dates, with two models simply wrong. The live-web fact-check settled it against Umm al-Qura tables.
What is the Gregorian date corresponding to 1 Ramadan 1448 AH according to the Umm al-Qura calendar of Saudi Arabia, and what day of the week will it fall on?
8 February 2027, a Monday — a direct, unhedged answer straight from the Umm al-Qura tables.
Monday, 8 February 2027 — adding that communities following physical crescent sighting may start a day earlier or later.
Monday, 8 February 2027 — the only panelist to pull live calendar pages while answering, citing Umm al-Qura sources inline.
Sunday, 7 February 2027 — stated confidently as "the standard conversion", and a day off.
Monday, 15 February 2027 — after a long visible calculation that talked itself a full week past the date.
Each model then graded the others' answers without seeing names — and the review split too: three reviewers confidently declared 15 February "the correct answer", one argued for 14 February, and only the retrieval-grounded panelist put the 8 February answers on top. On the scoreboard, a wrong answer tied with the right ones.
Blind peer review rewards confidence and clarity — but by itself it cannot settle a matter of fact. That is exactly why the Consilium pipeline does not stop here: the verdict still has to survive the live-web fact-check.
According to the Umm al-Qura calendar used in Saudi Arabia, 1 Ramadan 1448 AH corresponds to Monday, 8 February 2027 [1][2][6]. An independent weekday cross-check confirms that 8 February 2027 falls on a Monday [7].
Umm al-Qura dates are calculated astronomically per that standard, rather than by physical moon sighting [5][8].
Every claim in the draft verdict was taken to the live web. The date the review had argued about was confirmed against independent Umm al-Qura tables, and two confident-sounding claims were removed because no retrieved source backed them — this is the actual change log:
Hijri-to-Gregorian conversion looks trivial and is famously error-prone for language models: calendar arithmetic, the Umm al-Qura algorithm, and moon-sighting variance all trip them up. A lone chatbot gives you one confident date — with no way to tell whether it is the wrong one. A panel makes the disagreement visible, and the live-web check resolves it with sources you can inspect yourself.
Cited in the verdict
Numbers reflect the panel's retrieval order during fact-checking; only some sources were cited in the verdict. 14 sources were consulted in total — including the Umm al-Qura annual reference — six of them without being directly cited.