HaikeLite MT03 review with measurements (3x XHP70, 4x 18650)


Well-known member
Dec 22, 2015
Disclaimer: The review sample was provided by HaikeLite free of charge


XHP70 has proven to be the go to emitter for high output soda can lights. This time we're taking a look at the new player in the flashlight market, the HaikeLite and its competetively priced floody lumen beast the MT03 Devourer. The tested light has cool white emitters, but there's also a neutral white option with CRI80 LEDs.

Features and manufacturer’s specifications
Battery: 4x 18650
LED: 3x Cree XHP70, cool white or neutral white
Waterproof: IPX8
Mode memory, except moon
Electrical side switch
AR lens, 99% light transmission
Tripod thread

Output specs
Maximum output: 8760 lumens for 2.5 minutes, stepdown to 5500 lumens for 2 minutes, stepdown to 2400 lumens
Other output levels: 2700/1000/150/10/0.6 lumens
Light intensity: N/A
Beam distance: N/A

Measured dimensions and weight
Length: 124.8 mm
Head width: 72.2 mm
Handle width: 50.2 mm
Weight: 793 g including batteries

User interface

The MT03 uses an electronic side switch with blue backlight.


From off:
Single click turns the light on at the previously used brightness (not moonlight which is hidden)
Long press activates moonlight
Double click activates the highest mode
Double click plus holding the second click activates and deactivates lockout

From on:
Single click turns the light off
Long press cycles modes (low low, low, medium, high, turbo)
Double click activates turbo (single click takes you bak to the previous mode)

The mode cycling could a bit faster. It takes 1 second to advance to the next output level.

If you access moonlight by a long press or turbo by double clicking, cycling modes with a long press doesn't work. From moonlight accessed this way you can access turbo by a double click though. A single click takes you back to moonlight.

If you accessed turbo by a double click when the light is on, a single click will take you back to the previous mode rather than turn the light off. If you double click while on turbo you accessed with mode cycling, it will not change the output but a single click will not turn the light off but rather stay to turbo.

User manual: page 1, page 2







The light is packaged in a plain cardboard box with foam inserts and comes bundled with:

Cool lanyard
4 spare o-rings
User manual





PLENTY of lube around the threads and hefty double o-rings.



The battery tube houses 4 18650 cells in a 2S2P configuration. You have to insert the tailcap just right to align the different sized posts to their respective holes on the tube. A cross pattern for the batteries would have made screwing in the tailcap a bit faster.


The switch cover is a thick piece of rubber which is held on by a steel ring. In my testing the cover came off during the first turbo runtime when hot air rushed out of the light. After that the rubber piece came off very easily. In production lights the cover is glued (or not if you so prefer), but I'd prefer threads to make modding easier.




Beamshot comparisons



Ceiling bounce test after the lights had been used for ~10 minutes taking the other beamshots.


Moonlight mode is comparable in intensity to the Zebralight H600Fd MK III Plus's L1 mode.

Beam and tint
The shallow reflectors and big XHP70 LEDS make the beam very wide and floody. There is a discernible hotspot, but it is big and the intensity difference between the hotspot and spill is not huge like on a dedicated thrower.

There's a crosshatch pattern in the center of the beam which is visible on a flat surface. This would have probably disappeared had the emitters been rotated a bit like in the Olight X7.



Tint in different parts of the beam.


Tint in different brightness modes. Despite PWM there's a slight tint difference between modes. Probably because the frequency is so high (31 kHz).

Spectral data and color rendering

For spectral information and CRI calculations I use an X-rite i1Pro spectrophotometer with HCFR, Babelcolor CT&A and ArgyllCMS spotread for the graphs and data. For runtime tests I use spotread with a custom script and an i1Display Pro because it doesn’t require calibration every 30 minutes like the i1Pro.

Explanation of abbreviations (click link to read more)

CCT = correlated color temperature, higher temperature means cooler (bluish)
CRI (Ra) = color rendering index consisting of 8 different colors (R1-R8), max value 100
CRI (R9) = color rendering index with deep red, usually difficult for led based light sources, max value 100
TLCI = television lighting consistency index, max value 100
CQS (Qa) = Proposed replacement for CRI, RMS average of 15 color samples
CRI2012 (Ra,2012) = Another proposed replacement for CRI, consists of 17 color samples
MCRI = Color rendering index based on the memory of colors or 9 familiar objects
NEW Read more about the IES TM-30-15 method here (link is external)
TM-30 = The newest color rendering method using 99 samples. Preferred for comparing LEDs.
TM-30 (Rf) = Accuracy of colors, fidelity index. Replaces CRI(Ra).
TM-30 (Rg) = Gamut of colors, saturation index. Higher number means more saturated colors.
Tint dev. (“Duv” in the CTA screenshots) is the tint’s distance to the black body radiator line in the CIE graphs. The higher the number, the greener the tint. 0,0000 means absolutely neutral white and negative numbers mean rosy/magenta tint. Anything over 0,0100 can be described as visibly green.


HaikeLite MT03 spectral graph.


CRI Data


CRI and color comparison. Higher numbers are better, except for CCT (correlated color temperature) and Tint dev.

Runtimes and output

Please note: lumen measurements are only rough estimates
My diy 40 cm integrating styrofoam sphere has been calibrated using Olight R50 on that has with valostore.fi measuring it at 1178 lumens with their Labsphere. Results may vary with especially floody or throwy lights.

The light tested here has a 2S2P battery configuration, but there will also be a 4S version of the tailcap and driver with regulated output.

The runtimes for the HaikeLite MT03 were measured using protected Samsung 35E high capacity cells. They were chosen to make the comparison to the previously reviewed Olight X7 fair. Maximum output was also measured with higher drain unprotected Samsung 30Q batteries. See graps later.




Even though the lumen output is just an estimation, the comparison between the Olight X7 and the HaikeLite MT03 is fairly accurate since they both have a similar beam pattern with identical emitter configuration. The HaikeLite does offer more output for the first 2.5 minutes before it steps down to ~5000 lumens. The stepdown happens always at this time, since there's no continuos thermal regulation.



The light steps down at 2.5 minutes and a second time at a bit under 5 minutes even when cooled. The second stepdown can be avoided if the cooling is powerful. With my usual small USB powered fan the second stepdown reduced the output to about 2200 lumens. If I improved the cooling by lowering the ambient temperature and positioning the fan closer the light it never stepped down, just slowly declined til the end as the battery voltage dropped. This also for some reason also improved total output in lumen hours. The efficiency seems very good, since there is no boost driver needed for totally flat regulation.

Before every stepdown, the light makes a tiny whiny noise. Also with a 30Q battery it suddenly increases output at 25-40 seconds depending while also making that sound. With lower output cells the output declined steadily between stepdowns.

There's a LED indicator under the rubber switch cover that tells you when the batteries are getting low. Blue light flashes first, red red light when critically low. Just before the light shuts off and the low voltage protection kicks in, it flashes the main LEDs a couple of times.

The internal battery protection didn't activate during my runtimes. I measured a voltage of 2.64-2.68 from the batteries after the light had shut itself off.


The light behaves a bit differently on different batteries. I tested the turbo output for 10 minutes on a pack of Samsung 30Qs, protected Samsung 35Es, protected Keeppower 3500s (GA inside). Maximum output didn't differ greatly, but the behavior on the 30Qs is unusual.

On the 30Qs the average current draw for the first 2.5 minutes before stepdown was approximately 27 amps. So 9.1 amps per LED or 6.8 amps per battery. The current draw per battery also explains why a pack of four protected Keeppower 3500 mAh batteries (with a Sanyo NCR18650GA inside) has no trouble achieving maximum output of over 10000 lumens at startup and doesn't trip the over current protection (~8 amps).


It is possible to reset the turbo mode just by double clicking the switch. Just don't do it without proper cooling.

Standby current is only 33 µA so there's no real harm in forgetting the bateries inside the light for prolonged periods.


PWM is used on all modes except turbo before it steps down. The switching frequency is very high at 31 kHz so it is not visible to the naked eye but can manifest itself as scanlines in photos or video. Moonlight mode uses a lower frequency of 8 kHz, but that isn't visible in normal use either.





I measured and compared the surface temperature on a thermal camera from three similar lights. All lights were on their highest output mode and are shown here at 1 minute intervals.

At 1 minute

At 2 minutes

At 3 minutes

At 4 minutes

At 5 minutes

Here I reset the turbo mode from all lights by double clicking (two double clicks for the Olight to enable Turbo S). Temperature at 6 minutes.


At this point the HaikeLite is too hot and starts emitting faint smoke from the head and I abort the test. Forcing the turbo mode without effective cooling is not advised as there is no realtime PID temperature controller.

In normal use the HaikeLite does not get too hot. On turbo without a fan the head gets to 52°C before it starts to step down at 2.5 minutes gets to about 53°. On the other hand, as can be seen from the thermal images, the handle stays cooler, which is good for the skin but not so good for transferring the heat off the emitters. Olight seems to distribute heat more evenly than the HaikeLite or the MecArmy.


The HaikeLite is an impressive product for the price ($129). Its great mass and effective heatsinking allows very high output in short bursts and respectable levels (~5000 lm) even for prolonged periods if the light is properly cooled. The lack of sophisticated thermal regulation means that there is a possibility of damage if running on turbo continuosly. Never leave the light unattended!

The reflector arrangement is good for a floody light. There's no light falloff between the spill and the hotspot, rather the beam smoothly increases in intensity towards the hotspot. I would have liked to see the emitters angled differently to hide the donut hole, which is barely visible on a white wall on lower outputs.

The electronic switch's rubber cover is probably the weakest point on the MT03. It is too mushy and stiff, which makes activating turbo with a double click much harder than it has to be. I would also recommend switching to threads on the steel ring that holds the rubber cover in place, since unless it's glued, it falls off very easily and risks water getting inside the electronics.

+ High quality craftmanship
+ Very high output for 2.5 minutes
+ Useful moonlight mode
+ Good heatsinking with high mass
+ Very high efficiency on high (3500 lm)
+ Floody beam without dead spots
+ Low voltage protection at above 2.5 volts
- The tint shift throughout to beam typical for the XHP70 can't be avoided (where are all the MT-G2 multiemitter lights?)
- Stiff switch is hard to double click and the rubber cover with the press fit steel ring falls off easily (glued in production lights but available unglued if required for moddin purposes)
- Repeated physical lockout of the head may also unscrew the driver retaining ring making the light not work
- Thermal transfer to the handle is not great
- Repeated turbo activation without cooling causes overheating
- Timed stepdown instead of true realtime thermal regulation


Well-known member
Jan 1, 2016
Well done,I like your thermal images the best. I am liking the design of this light more than the other pop cans.