In May, the Air Force was planning to pay $2.97 billion for 42,275 Maverick anti-tank missiles in fiscal years 1984-1988. Revising the plan in September, the Air Force foresaw paying the same $2.97 billion -- but for 12,025 fewer Mavericks.

In addition, the Air Force calculated in May that the 490 initial production Mavericks it wanted to buy from Hughes Aircraft Co. would cost $480,000 each. In September, when it was seeking only 200, the unit cost was $1.1 million -- more than twice as much.

The drastic changes are disclosed in an Air Force "quantity/cost" analysis, one of several Pentagon documents obtained by the Project on Military Procurement. The project, which monitors military programs, released the documents to several reporters for publication this morning.

The Air Force provided an explanation in a Pentagon report to Congress on increases in the estimated costs of 39 major weapons in the quarter ending June 30. The report said the cost of Maverick grew from $5 billion to $6.2 billion -- the largest growth among the listed 39 weapons. The unit cost rose 23 percent.

The Air Force attributed 49 percent of the $1.2 billion increase to a slowing of planned procurement, 27 percent "to a correction in the way the contractor calculated inflation in prior years (FY 75-FY 81)," 18 percent to "receipt of contractor quotes that exceeded the original estimates," and 6 percent to "estimating error and changes in support/training equipment."

The Air Force prepared the documents for a weapon review panel, led by Defense Under Secretary Richard D. DeLauer, which advised approval of "low rate pilot production" of 200 "smart" heat-seeking missiles. Secretary Caspar W. Weinberger accepted the advice Sept. 29.

In a decision memo to the Air Force, Weinberger said he was releasing $160.2 million in fiscal 1982 funds for the "Segment I" production of the 200 Mavericks. The memo did not mention the $61 million released earlier for the same batch.

He also said his "acquisition strategy" was to limit fiscal 1983 "Segment II" production to 43 a month, subject to a review next Feb. 1 of the results of further tests of experimental Mavericks. By contrast, the Air Force had proposed much faster acquisition, starting with a fiscal 1983 procurement rate 2 1/2 times higher.

The Pentagon said the initial 200 missiles will enable the Air Force to continue tests for "reliability validation" and to do follow-up testing and evaluation "under operational conditions." The Pentagon said that it will approve mass production only if Maverick "continues to prove its ability as a cost-effective tactical weapon system" at night, in "limited adverse weather," and in "battlefield haze."

Critics say that five years of testing have failed to show Maverick to be effective in realistically simulated combat conditions even in the daytime, particularly in comparison with a relatively cheap, mass-produced 30-millimeter cannon of proved reliability that destroyed or mobilized all of the 10 Soviet main battle tanks in an operational test.

In March, Anthony R. Battista, the House Armed Services Committee's staff expert on Maverick, warned that the missile "would probably get more pilots killed than they would kill targets." He charged that approval of the 200 initial Mavericks would lock in the ultimate purchase of 61,000.

In July, the General Accounting Office urged the Pentagon to hold off on the initial buy, saying that five years of operational testing failed to show that Maverick "can be used effectively by U.S. military personnel in combat." It did poorly even when pilots were alerted to "what to look for" and in other "very favorably test conditions," the GAO said.

Starting in February, the Air Force did operational tests of nine Mavericks. According to a report last summer, the first eight were found to have a less than 30 percent probability of working properly after being airborne for 14 hours, although the target standard was 85 percent reliability.

Martin F. Chen, the Air Force acting assistant secretary for research and development, said last night, "It is true that we have not reached the 85 percent threshold." At the same time, he said, "we have had an increase in the trend toward reliability improvements." Chen confirmed that two of the nine shots were unsuccessful but said that seven were direct hits.