In 1919 the Bhandarkar Oriental Research Institute at Pune began an attempt to produce a critical edition of the Mahabharata. V.S. Sukthankar became general editor in 1925 and worked on it until his death in 1943; S.K. Belvalkar and then P.L. Vaidya carried it on, and the completed edition was announced in September 1966. It took forty-seven years, collated 1,259 manuscripts in a dozen scripts, and filled nineteen volumes.
What it produced is a constituted text of about 89,000 verses, appreciably shorter than the received vulgate, printed above an apparatus that records what each manuscript family has, with disputed passages marked and rejected material relegated to appendices. The editors' rule was conservative: where the northern and southern recensions diverge, prefer the reading that best explains the others, and treat the Kashmiri Sharada manuscript as the most reliable single witness.
The distinction that matters is what such an edition can claim. It recovers an archetype, the earliest state of the text that the surviving manuscripts allow reconstruction of, which Sukthankar himself dated to a few centuries CE. It does not recover an original poem, and Sukthankar said so. Later critics have argued that privileging the Sharada codex builds a northern bias into the result, and that a work which grew by performance may never have had a single archetype to recover. The edition nonetheless remains the basis of serious scholarship, and of the Chicago translation begun by J.A.B. van Buitenen in 1973 and still being completed by other hands decades later.
For general readers the older complete English version by Kisari Mohan Ganguli, published between 1883 and 1896 from the vulgate, is still the one most often quoted, which means the most widely read English Mahabharata is not the critical text.