THE PENTAGON has reached a point in the defense buildup that administration officials used to deny would occur. It is increasingly clear that the buildup has been underpriced. The defense budgets now projected, large as they are, will not buy all the weapons the planners earlier said they would. The services are having to choose; there is growing pressure on the weaker weapons that they want. A classic example is DIVAD, the Army's new radar-controlled divisional air defense gun.

To accompany its tanks in battle, the Army wants 614 of these guns (the current cost estimate for which is $4.2 billion), and has already ordered 146, of which 50 have been delivered. The orders were placed before the weapon was fully tested, under a procedure intended to save time. The procedure would have been fine had DIVAD then passed the tests. Unfortunately it has not. The radar has been misled, the gun has not been quick enough and it has jammed.

Last year the administration requested and Congress approved purchase of the next 117 DIVADs, but Defense Secretary Caspar Weinberger took the unusual step of holding up the expenditure, pending further testing. That has now been done, and Mr. Weinberger is scheduled to announce a buy-or-no- buy decision before Sept. 30, the end of the fiscal year.

The Army and DIVAD's manufacturer, Ford Aerospace, a division of Ford Motor Co., both say the gun did well in its further tests this year. The most dramatic though not necessarily most important of these was a live-fire exercise in June. Ford has put about a paper saying that in this exercise the weapon "destroyed six of seven high-performance aircraft and three of three helicopters presented." But a cochairman of Congress' self-styled Military Reform Caucus and a critic of DIVAD, Rep. Denny Smith (R-Ore.), says this is misleading. It sounds as if DIVAD was offered 10 targets in realistic circumstances and shot down nine. In fact weapons experts say such tests tend to be less realistic than they sound, because of the cumbersomeness of the drone aircraft involved. There were also many more than 10 fly-bys in the test -- 32 "presentations," the Army says -- and not all the claimed kills are clear. The test used proximity rounds, which are meant to disable planes by bursting near them. The three helicopters were felled by these bursts, the Army says, and two of the fixed-wing aircraft. The other four fixed-wing targets that Ford said were destroyed were actually "command destructed" by the range safety officer; he blew them up when the proximity rounds did not bring them down right away. No one knows whether the bursts from the rounds themselves would have brought them down eventually.

The Army already has several other weapons to protect forward units against air attack (and the Air Force has still more; that is one of its main jobs). But the Army says the weapons it has are not good enough to meet the envisioned threat, and that it will need either DIVAD, on which it has already spent about $1.5 billion, or a comparable weapon. Spokesmen also warn against holding DIVAD or any complex modern weapon to too high a standard too early in its life cycle; they speak of teething problems.

The trouble with these kinds of arguments is that they are open-ended; they leave no way to say no. DIVAD is a costly weapon, the need for which is not completely clear and in whose performance no one can have great confidence on the record so far. The Army may need more protection for its tanks, but Mr. Weinberger needs even more to impart a sense of discipline and credibility to the procurement process. A decision to keep buying DIVAD would take him in the opposite direction. DIVAD has become a symbol. If it can pass muster in its present state, the message is that any weapon can.