UltraFire 9T6 (with 9 x XM-L) Review (2-3x 18650 or 26650)


Flashlight Enthusiast
May 5, 2010
Hove, UK
The UltraFire 9T6 is the second one up for this set of three reviews of DX lights. Quoted as having 7200lm output, this is the highest output LED light listed on DX. With a 65W HID I have measuring in at 3560lm, 7200lm is a tall order, so how close will it get?

(The first review in this series can be found here.)


Initial Impressions:

Much as we should not judge a book by its cover, the UltraFire 9T6’s packaging is a let down. The lack of proper protection meant that the extension tube was dented out of the box. Although I was able to ‘ease’ this back into shape, it is not a good start.

Moving on from this initial disappointment the 9T6 is a substantial light. As the battery tube takes 26650s the tube diameter is equivalent to one that takes C-cells, making this more of a handful than the more common 18650 size tube. Somehow it doesn’t seem right not to include the extension tube and once in place the 9T6 takes you back to the days of the 4 C-cell Maglite, but with substantially different performance!

Even beyond the 7200lm quoted by the manufacturer (and so listed by DX), the 9T6 claims 9500lm on the side of the battery tube. 9500lm! – better just ignore that; it is a pity that a more realistic claim is not made for the output.

Still, it has 9 XM-L emitters, this should be interesting.


What is in the box:

The outer packaging.


Inside the box the components are wrapped in a thin bubble wrap.


During transit, the package has been dropped and due to the lack of padding, the end of the extension tube hit the head of the light denting the tube wall inwards.


In the box are the main light head and battery tube, the extension tube, tail cap and plastic tubes to allow 18650 cells to be used.


Taking a closer look and looking inside:

So, the main reason you might want this light is the massive output that is claimed. To achieve big output, this light has a big head with 9 XM-L emitters in a compound reflector.


Looking dead on into the lens, the reflector looks a bit like a few soap bubbles clumped together. Each of the outer ring of 6 emitters has a reflector with alternately merges with one or two of the inner three emitter’s reflectors. There is a pattern, but it looks slightly dishevelled.


Broken down into the main components.


The cooling fins.


The graphics are printed onto rather than etched into the surface.


Threads look asymmetrical, but are cut cleanly enough.


The positive terminal in the head is a large spring.


The tail-cap contact is a large brass spring loaded plunger. A standard reverse clicky switch.


The tail-cap switch boot is made from GITD material.


After repairing the dent to the extension tube, the surface finish has a few fine cracks visible. This surface finish looks to be paint or lacquer rather than anodising.


The 9T6 fully assembled.


Modes and User Interface:

The UltraFire 9T6 has a typical five mode driver with memory. The switch is a reverse clicky action.

Click-on to the last used mode, then half press the switch briefly to change mode. Modes are High, Medium, Low, Strobe, SOS back to High etc.

As long as you leave a second or two between turning it off and on again, it will remember the last used mode.

Batteries and output:

The UltraFire 9T6 has four possible battery configurations:
2x 18650 or 2x26650 – without the extension tube
3x18650 or 3x26650 - with the extension tube

The protected 26650 cells available on DX seem a little longer than even the longest protected 18650s so when loaded with 3 cells, the tailcap does not screw all the way down, and it is easy to dent the end of the cells slightly. As the threads are bare, there are no issues with operation, but even with the springs in the positive and negative terminals, there is not enough movement to fully accommodate 26650s

26650 cells are a good fit in the battery tube, but even with the adapters provided, 18650s are a loose fit. However having these more common cells as an option gives far greater flexibility.


So now down to the crux of the matter – just how bright is this?

To measure actual output, I built an integrating sphere. See here for more detail. The sensor registers visible light only (so Infra-Red and Ultra-Violet will not be measured).

Please note, all quoted lumen figures are from a DIY integrating sphere, and according to ANSI standards. Although every effort is made to give as accurate a result as possible, they should be taken as an estimate only. The results can be used to compare outputs in this review and others I have published.

The output measurements were made using 26650 cells.

Starting off with 2 x 26650 (and also confirmed with 2x 18650)

UltraFire 9T6 output mode 2 x 26650I.S. measured ANSI output LumensPWM frequency (Hz)
High (after cleaning the threads and contacts)2330

Whoa! A 9 XM-L light claiming 9500lm giving 233lm on high. Initially I thought this was a fault, but no, it is exactly what this light gives on 2 cells.

Better move swiftly along now to use the extension tube and load up 3 x 26650s

UltraFire 9T6 output mode 3 x 26650I.S. measured ANSI output LumensPWM frequency (Hz)
High (after cleaning the threads and contacts)3717185

Strobe frequency is 9.6Hz

Peak current draw on high with 3 x 26650 cells (12.47V) is 2.28A

Well, that is more like it. 3717 ANSI lumens available instantly (beating a 65W HID). So not the 7200lm or 9500lm claims, but 3717lm is still a lot from a production UltraFire light. Not bad at all.

For the first runtime graph the 26650s from DX were used. During the runtime, the head reached 46º C under a strong cooling fan, so a lot of heat is being dissipated.


The output does steadily decline, roughly halving after only 28minutes, continuing to decline until it hits about 500lm after an hour of constant running. No evidence of any regulation.

After running the 26650 runtime trace I decided to try it on good quality 18650 so loaded up three of AWs 3100mAh cells.

The results were strange with a lot of flickering.


Thinking this was some sort of contact issue I cleaned all the contacts again (as I always check and clean contacts before runtime tests) and ran it again. This time it was even more uneven, though had a slightly better total output over the test period.



So I ran it again with Xtar 3100mAh cells. The 2.28A draw should not cause problems with either the AWs or the Xtars.

The Xtar trace seems a bit better and the point at which it reaches the 2000lm output level there may have been some driver interaction causing the fluctuating output – or not.


I’m not sure if this odd behaviour on 18650s is the cell chemistry, the protection circuitry. Or maybe it was the effect of the loose fit of the 18650 cells (although the test was run in a fixed rig) combined with the effects of heating during the runtime test, or maybe the barely perceptible vibration from the cooling fan was the reason for these results.

Back on the 26650s the output is smooth as in the first trace.

In The Lab

NEW for Winter 2012 ANSI standards include maximum beam range. This is the distance at which the intensity of light from an emitter falls to 0.25lux (roughly the same as the lux from a full moon). This standard refers only to the peak beam range (a one dimensional quantity), so I am expanding on this and applying the same methodology across the entire width of the beam. From this data it is possible to plot a two-dimensional ‘beam range profile’ diagram which represents the shape of the illuminated area.

In order to accurately capture this information a test rig was constructed which allows a lux meter to be positioned 1m from the lens and a series of readings to be taken at various angles out from the centre line of the beam. As the rig defines a quadrant of a circle with a radius of 1m, all the readings are taken 1m from the lens, so measuring the true spherical light intensity. The rig was designed to minimise its influence on the readings with baffles added to shield the lux meter from possible reflections off the support members.

The distance of 1m was chosen as at this distance 1lux = 1 candela and the maximum beam range is then calculated as the SQRT(Candela/0.25) for each angle of emission.

In this plot, the calculated ANSI beam ranges are plotted as if viewed from above (for some lights there may also be a side view produced) using a CAD package to give the precise 'shape' of the beam.

Starting with the 5m range grid you get an idea of the broad flood of light blotting out the range grid almost entirely.


Then zooming out to the 50m range grid showing the extent of the wide beam’s range. Even with the wide flood of light, the ANSI beam range still reaches 300m.


The beam

The indoor beam shot shows how blinding this light is. At close range 3700lm is way too much! There are multi-emitter artefacts at the outer edges of the spill.


Now going outdoors, to put things in perspective, just as I did with the previous review, this is the TK41 (my frequently used reference light due to its well-known excellent performance).


And on the same exposure setting, here is the 9T6 on full blast. That is 3700lm for you!


What it is really like to use…

As mentioned in the initial impressions, the form factor of this light takes you back to the days of multi-cell Maglites. Long wide battery tube, combined with a reasonably heavy construction makes this a light you know you are carrying.

It is just not worth bothering to only use two cells in this light. It would have to be an emergency when you really only have two cells available as the performance on 2 cells is worlds apart from the output on 3 cells.

This light is suited to massive output for short periods of time. As shown in the runtime graphs, you only get 15minutes or so of 3000lm-and-above output. Not so much a limitation of the 9T6 itself, more on drawing 2.3A from each cell and the resulting drop in voltage. The 9T6 wants all 12.6V which it will not get for long.

The PMW is present in all output modes and at 185 Hz is not too noticeable.

Somehow not using high seems like cheating the UltraFire 9T6 out of giving its best, but 3700lm is too much to be practical for regular use – unless you need to provide floodlighting for a sports field.

The ‘up to 9500lm’ claimed on the light itself is not going to happen, and the manufacturers claim of 7000lm (hence DX quoting this figure) also overblown, but as the output tests showed, this is an actual ANSI output of 3717lm which is not to be sneezed at!

Solid construction and very power hungry, the UltraFire 9T6 does not disappoint as far as awesome output goes - just have a ready supply of high capacity cells on hand to feed it.

Is it practical? – the jury is still out on that….. but is it fun? – Yes 3700lms worth.


Test sample provided by DX for review.

(Note – prior to posting this review in the main ‘flashlight reviews’ forum, the CPF site moderators confirmed that this was correct forum)
Last edited:


Flashlight Enthusiast
Dec 8, 2007
Why didn't they just use 7x xm-l2's?
If it isn't going to have regulation, maybe 32650's would be more appropriate?


Flashlight Enthusiast
Sep 4, 2009
The 5*XM-L version will match that for output with a $25 driver upgrade.

I'd be very interested to see the results if you were to install a KD "beast" driver in this one!


Newly Enlightened
Feb 21, 2013
I got the 12*XML version (claiming 13,000 lumens) for a present, and it sure is bright, but I'd guess more like 4,500 lumens. When it rains, the driver appears to turn LEDs off/on even on high, I get the "cartoon rain" effect that I've noticed on med and low power using my 3*XML. Anyone have any stats/review/links on a 12*T6 XML Trustfire?


Newly Enlightened
May 9, 2013
Great review! Huge relief to finally see a credible source test the actual lumen output of this specific light. 3717 lumens on eBay currently for as cheap as $67 "buy now"

there is a nearly identical light as this, also on eBay, with 12 total Cree XML T6 LEDs ($80.38, buy now) instead of 9. Has it been tested yet?

Barrel mount

Newly Enlightened
Jun 14, 2013
I've a 9xT6 branded Trust-fire. From your close up pictures, design fins and tube are identical, making me think Trust-fire and Ultra-fire are cousin companies or at least ate Thanksgiving dinners on the same table. LOL. Only the reflector is different.

I run a small biz and close up at night, so this 9xT6 makes me feel better with its intimidating delivery of light. My G29 would come in to play if the light doesn't work.

This afternoon, I tried out a 9x board from Kai, supposedly would do 75w plus. It only flicker, briefly on for less than a second then off. The 26650 5,000mAh batts, three of 'em, don't have sufficient current for this board. I think the OEM board would do something like 25w. So I pulled out a PS oldie from RadioShack spec'd at 12v 2.5A. Immense light, noticeably brighter than OEM, but trips circuit board after about 15sec. Kai claims it can do closer to 9A!

in practical sense of mobility, the OEM would do just fine. I did upgrade supply leads to larger gauge, though.

My point is that stuff more than 4,000lm is impractical for mobility. Those 26650s just can't provide enough current delivery. This means 12x designs are only marginally brighter than 9x, since you're limited by batteries.
Last edited:


Mar 17, 2011
Subwoofer, if you still have this, would it be possible to do a runtime graph on medium?

Thank you,



Flashlight Enthusiast
May 5, 2010
Hove, UK
Subwoofer, if you still have this, would it be possible to do a runtime graph on medium?

Thank you,


I do still have it. I actually use it for ceiling bounce lighting for some of my photography. I'll add your request to the list, but don't hold your breath ;-)


Mar 17, 2011
Thanks and no worries :)

What would you guess it does? 2 hours above 1000 lumens possible on medium?


Flashlight Enthusiast
May 5, 2010
Hove, UK
well, DX now have a 15x XM-L light! gonna test it? :)

Actually they have two! One by Nitefire and one by Ultrafire.

Both take 4x26650/18650 which doesn't add up to me as far as power source goes. This one with 'only' 9 dropped in output very fast with the three cells (3 LEDs per cell), so 15 LEDs drawing from four means a very basic comparison of 3.75 LEDs per cell, so the cells are loaded more heavily. I suspect this will make the light a 'flash in the pan'. Might be fun to have a closer look though.

Latest posts