The conditions to run the test were established prior to the day shift of 25 April 1986. The day shift workers had been instructed in advance and were familiar with procedures. A special team of electrical engineers was present to test the new voltage regulating system. As planned, on 25 April a gradual reduction in the output of the power unit begun at 01:06 a.m., and by the beginning of the day shift the power level had reached 50% of its nominal 3200 MW thermal. At this point, another regional power station unexpectedly went off-line, and the Kiev electrical grid controller requested that the further reduction of Chernobyl's output be postponed, as power was needed to satisfy the peak evening demand. The Chernobyl plant director agreed and postponed the test.

At 11:04 p.m., the Kiev grid controller allowed the reactor shut-down to resume. This delay had some serious consequences: the day shift had long since departed, the evening shift was also preparing to leave, and the night shift would not take over until midnight, well into the job. According to plan, the test should have been finalized during the day shift, and the night shift would only have had to maintain decay heat cooling systems in an otherwise shut-down plant; the night shift had very limited time to prepare for and carry out the experiment. Further rapid reduction in the power level from 50% was actually executed during the shift change-over. Alexander Akimov was chief of the night shift, and Leonid Toptunov was the operator responsible for the reactor's operational regimen, including the movement of the control rods. Toptunov was a young engineer who had worked independently as a senior engineer for approximately three months.

The test plan called for the power output of reactor 4 to be gradually reduced to 700–1000 MW thermal. The level established in the test program (700 MW) was achieved at 00:05 on April 26; however, because of the natural production in the core of a neutron absorber, xenon-135, reactor power continued to decrease, even without further operator action. And as the power reached approximately 500 MW, Toptunov committed an error, inserting the control rods too far, bringing the reactor to a near-shutdown state. The exact circumstances are hard to know, as both Akimov and Toptunov died from radiation sickness.

The reactor power dropped to 30 MW thermal (or less)—an almost completely shutdown power level that was approximately 5 percent of the minimum initial power level established as safe for the test. Control-room personnel therefore made the decision to restore the power and extracted the reactor control rods, though several minutes elapsed between their extraction and the point that the power output began to increase and subsequently stabilize at 160–200 MW (thermal). In this case the majority of control rods were withdrawn to their upper limits, but the low value of the operational reactivity margin restricted any further rise of reactor power. The rapid reduction in the power during the initial shutdown, and the subsequent operation at a level of less than 200 MW led to increased poisoning of the reactor core by the accumulation of xenon-135. This made it necessary to extract additional control rods from the reactor core in order to counteract the poisoning.

The operation of the reactor at the low power level with a small reactivity margin was accompanied by unstable core temperature and coolant flow, and possibly by instability of neutron flux. The control room received repeated emergency signals of the levels in the steam/water separator drums, of relief valves opened to relieve excess steam into a turbine condenser, of large excursions or variations in the flow rate of feed water, and from the neutron power controller. In the period between 00:35 and 00:45, it seems emergency alarm signals concerning thermal-hydraulic parameters were ignored, apparently to preserve the reactor power level. Emergency signals from the Reactor Emergency Protection System (EPS-5) triggered a trip which turned off both turbine-generators.

After a period, a more or less stable state at a power level of 200 MW was achieved, and preparation for the experiment continued. As part of the test plan, at 1:05 a.m. on 26 April extra water pumps were activated, increasing the water flow. The increased coolant flow rate through the reactor produced an increase in the inlet coolant temperature of the reactor core, which now more closely approached the nucleate boiling temperature of water, reducing the safety margin. The flow exceeded the allowed limit at 1:19 a.m. At the same time the extra water flow lowered the overall core temperature and reduced the existing steam voids in the core. Since water also absorbs neutrons (and the higher density of liquid water makes it a better absorber than steam), turning on additional pumps decreased the reactor power still further. This prompted the operators to remove the manual control rods further to maintain power.

All these actions led to an extremely unstable reactor configuration. Nearly all of the control rods were removed, which would limit the value of the safety rods when initially inserted in a scram condition. Further, the reactor coolant had reduced boiling, but had limited margin to boiling, so any power excursion would produce boiling, reducing neutron absorption by the water. The reactor was in an unstable configuration that was clearly outside the safe operating envelope established by the designers.