Allen Key Size Chart: Metric & SAE Set Table (Free PDF)

The table above doesn’t describe the screws driven by these allen keys; it’s about the allen keys themselves, or more exacty L-shaped hexagon bars as described in the standard ISO 2936:2014

You will find your key in the Key size s (mm) column. The number there is what is stamped onto the key itself, and that is also the distance between the opposite sides of the hexagon bar (the width across the flats). The next column over (Decimal equivalent (in)) convert the nominal value into inches. This allows you to compare it to the SAE drawer at a glance.

The other columns are the length of the two arms of the hex and the manufacturing limits. Over on our Hex Key Size Chart in this same category is where you’ll find which key goes with which screw. This page is all about the tool in your hand.

From Table 1 of the standard, the Across flats max and Across flats min columns represents the tolerance band. To ensure every key fits into its intended socket, each key is intentionaly finished just under nominal to allow for the socket being machined slightly over nominal.

A 6 mm key should of been between 5.95 and 6.00 mm; a 10 mm key should be between 9.94 and 10.00 mm. That’s what makes a new key feel tight and an old key rock back and forth. It’s only a few hundredths of a millimetre of working clearance and when wear gnaws away at it, the key begin to round the sockets. Retire any key measuring below the min column across its flats.

L. Long arm l1. Available in three series. The classic hardware store key is standard; long provides added leverage and reach; extra long allows the maximum swing with the short arm engaged or when working in a deep recess or down into something like a screw at the base.

Long and extra long series is defined in 1.3 to 18 mm. Only available as standard lengths in 0.7 and 0.9 mm, which doesn't have an equivalent long or extra long arm. Everything from 19 mm up is standard only. These are the nominal maxima for length. By ISO 2936, the long arm can be made short by as much as 2 mm on the smallest keys and 12 mm on the largest.

Short arm l2, This is the stubby leg you stick into a socket for high torque applications. The long arm becomes the lever.

The odd ones, you check: 3.5, 4.5, 7, 9, 11, 13, 15, 16, 18, 21, 23, and 29 mm (the whole ISO table), but the trade only stocks a fraction of those. Nearly all of what’s available for sale is 1.5 to 10 mm plus 12, 14, 17 and 19 mm.

Then there’s the row of 1.3’s. It measures at most 1.27 mm across the flats, which is why manufacturers like Bondhus rate the same-sized tool as 1.27 mm. The 1.27 mm and 1.3 mm keys aren’t neighbours. They’re the same size. You don’t mix them up with the 1.27 mm (or 0.050 in, nominal) of the SAE system. Here the systems actualy overlap.

Not all metric hex tools fall under this table. The most popular sizes have the same width across flats based off both standards (ASME B18.3.2M, the American metric key standard), but the arm lengths are a little different. It skips the half sizes altogether and uses 32/90 mm for a 6 mm key compared with 38/96 mm here.

There’s also a whole set of length conventions for 1/4 in hex insert bits, T-handles, and screwdriver-style hex drivers. Tools with anti-camout profiles like Wera Hex-Plus still stick to the across flats limits in this table; these gauge accordingly.

ASME B18.3 defines the dimensions for inch series hexagon keys in this table. This standard apply to both the hex keys and the socket screws they fits into.

The first column is the nominal width across flats, titled Key size (in). Fractional sizes starts at 1/16 inch and increase in fractions of an inch. Three decimal sizes makes up the bottom of the range: 0.028, 0.035, and 0.050 inches. There are no standard fractional names for these decimal sizes.

The next two columns is the tolerance band for the hexagon: across flats max and min. The mm approx column convert the nominal size to millimeters for comparison against the metric table. The final three columns are the arm lengths.

Our Hex Key Size Chart will tell you which key go into what screw. It's a table of the tool itself.

The ASME B18.3 specifies two L-key patterns, both with a matching short arm. Most come from a set containing the compact short series key. Long series keys is available and sold with an additional ~3 inches of reach on their middle-sized keys.

These maximum dimensions is all standard. Each dimension may runs up to 0.188 inches shorter. Largest keys can be as much as 0.5 inches less. A 1/4 inch short series key are between 3.156 and 3.344 inches long. Its short arm will range from 1.031 to 1.219 inches. Ball end version typically correspond to long series dimensions and are not defined in the standard.

The tolerance logic follow the metric table. Keys are finished under nominal so they enters sockets made over nominal. For example, a 1/4 inch key will measure between 0.2485 and 0.2500 inches.

Proportionally, there is a larger tolerance band on small keys. That’s part of why tiny keys seems to be sloppy in cheap screws. At that size, a few ten-thousandths of an inch make a difference.

Standard alloy keys has hardness specified as near Rockwell C 45 to 53. For sizes through 3/8 in, distributors quotes a minimum of C48. Above 3/8 inch, they quotes a minimum of C45.

Avoid these traps: The three decimal inch sizes can be hard to read correctly. 0.050 inch is 1.27 mm precisely, which makes it a metric 1.27/1.3 mm key. But 0.028 inch and 0.035 inch has no metric twin. Reduced fractions cause mispicks in a mixed drawer of keys.

Between 5/32 and 1/8 is 9/64. A lot of cheap sets leaves that one out. Between 3/32 and 1/8 is 7/64, an easy mistake.

Only according to the current standard are there keys larger than 1 inch made just in the short series. For example, Unbrako catalogs shows 1 3/4- and 2-inch keys for the biggest cap screws.

In an emergency, you might substituted metric for SAE by reading their near-equivalents table from our Hex Key Size Chart. Don't eyeball the mm column in this table.

Security keys, fold-up sets, hex bits and T-handles aren’t included here. Those all has lengths that are maker-specific.

And it doesn’t include metric keys. The sizes may appear similar, but 3.97 mm equals 5/32 inch, not 4 mm. Using one in the other is a classic technique for rounding off a socket. You should of checked the sizes actualy more carefuly.

Here are the maximum torque values at which each size of Allen key will fail when tested to destruction. The Torsional shear strength avg (in-lb) column come from an engineering document published by Lubker Distribution and MSC Industrial. This document details how they tested the torsional strength of ASME B18.3 inch keys.

Standard alloy steel keys failed during these destructive tests when torque was applied near the end of the long arm while holding the short arm inside a hardened socket adapter. The numbers indicates the average torque at which this occurred. The Nm approx column convert those numbers into approximate newtons-meters using a ratio of 0.113 Nm per in-lb. Keep these numbers in mind as a ceiling; don’t aim for them.

Now, don’t mistake these numbers for tightening specs. Key failure torques is much higher than screw seating torques. Take a 1/4-20 alloy socket head cap screw, for instance; it seats on a 3/16 in key at approximately 87 in-lb but fails on average at 342 in-lb on the key itself. That’s a healthy gap and is by design. It allows a properly sized, fresh key to strip or break the screw long after the screw is fully seated.

If you’re tearing up keys under normal working conditions, then the key is undersized for the screw, the key is shot or something else is preventing the screw from turning.

The other rating given by manufacturers is the torque rating, and it is lower. They are the working ratings of the manufacturers. One will be rated at 655 in-lb and another at 615 for the same size 1/4 in key. We printed the destruction average since all the sources agreed on this number.

Better keys go higher: Unbrako lists their minimum twisting shear as 20 percent over ordinary keys; and Bondhus claims much of the same for its Protanium steel. When a good key fails, it cleanly shears off square flush with the socket. Mushrooms or splinters? Then it was too hard or too soft.

ISO 2936 governs metric keys. It mandates that each size withstand a minimum test torque. ASME B18.3.2M also lists metric torque ratings. Those are numbers we didn’t print because we couldn’t cross verify them between two separate publicly available sources.

For a ballpark guess you can go down the mm column on this table. The 6 mm is right under 1/4 in, the 8 mm is right over 5/16 in, the 10 mm is right over 3/8 in, etc., so their capacities should roughly bracket.

But it also calls out a requirement of ductility in ISO 2936; up to a 14 mm key shouldn’t snap off like glass, but instead twist 60 or more degrees before breaking.

There are two caveats with this practical. One, these figures are based off hand force on the long arm. They also assume the short arm is fully engaged to socket depth. Putting a pipe over the arm multiplies leverage beyond the design conditions. People round sockets and shear off keys all the time this way.

Second, the figures are for straight hex engagement only. Ball ends break at much lower torque at the thin neck behind the ball. That’s what the next table is for.

The above chart is a feature comparison between the two types of tips (not sizes) on the ends of allens. From left to right: Straight hex end refers to the plain full-hex tip that both the ISO 2936 and ASME B18.3 standards dimension. Ball end refers to the sphere-on-a-neck type of tip on the long arm of most premium L-key sets.

What’s the difference? That’s the question for the last column, what it costs or gives you at the workbench.

That’s where it gets down to the magic number: the entry angle. Both Wiha and Bondhus say their ball ends enters a socket as far as 25 degrees from the screw axis. That’s why they’re built that way; in most cases, a straight shot isn’t possible because of a housing wall, a frame tube, a hose or something else. Even so, the ball end finds its way to the screw and turn it.

The problem is socket engagement. Rather than six flats gripping across the entire depth of the socket, there’s just a slim band encircling the middle of the ball touching those walls. When you get the tool at full tilt the contact area becomes even smaller.

The torque rule follows that geometry. Because the neck behind the ball is thinner than the hex bar, a ball end will fail far below its torsional limit in the torque table. That failure is quickest in small sizes where the neck are already fragile.

Makers use different ball ends that are not standard in any dimension, which leaves different amounts of residual capacity. Honest ones sell it as a speed and access feature, not a strength feature.

A safe working pattern is to loosen the screw first and then use just enough final seating torque with the short arm's straight portion. Only use ball end for fast mid-turns. That's why L-keys with ball end has a straight hex tip on their short arm.

Ball ends come in handy with assembly jobs like machine guards, furnitures, and bicycle frames. There are lots of screw and bad access.

If the socket starts to round, skip it. Use it on a corroded or thread-locked screw? Nope. On a tiny grub screw? Uh-uh. The neck will break off and get stuck in there. That's a two minute job that'll turn into an extraction.

If you are aiming for a lot of torque, the solution is to either use another type of fastener or reposition the screw over time. The answer isn't to get a harder ball. You should of picked a different tool.

Before purchasing a set, check out this table so you’ll know its contents, and also which screws you still won't be able to touch. Each row represent a single traditional format for a set. The columns are Pieces and Sizes included, which is the makeup most brands ship. Example products are names of confirmed existing sets matching the given run. Gaps to know about is the sizes the format doesn't include.

There are two primary runs: the 9 piece metric run (1.5, 2, 2.5, 3, 4, 5, 6, 8, 10 mm); that’s the industry standard. You’ll find this same set of sizes from companies like Bondhus, Wera and Wiha. From the smaller sizes it increment in 0.5 mm. Then it jumps from 6 to 8 to 10. That’s because there is almost no fasteners available for 7 or 9 mm.

There is also a 13 piece SAE run (0.050 in through 3/8 in), which again is the same between Bondhus, Eklind and Wiha. This is why you have 9/64 in in your drawer, just as an example. What kind of screws each of these runs covers is precisly what our Hex Key Size Chart lists. To put it simply, the metric nine reach up to M12 cap screws; the SAE thirteen reach up to 1/2 in.

The compromise between folding up versus staying unfurled is space. Fold-ups sacrifice capacity for pocketability; where they really differ is in composition. For instance, Bondhus folds up to 7 keys (2-8 mm), whereas Eklund do 7 (1.5-6 mm). They both has what you want. Eklind has a 1.5 instead of an 8, so it has your little set screw. Bondhus has an 8, so it has your M10 cap screw.

On the SAE side, there is only nine keys (they’re all about uniformity) going from 5/64- to 1/4-in., which is more like a medium-stretch set. Then there’s the way the handle operate. A fold-up will deliver good torque throughout its body, but can’t access recessed screws like the L-key drops straight into them. And then there’s just the lack of the short-arm high-torque deal.

Buy advice for the gaps column. No matter the set, you're buying singles if your work uses 9 mm or 7 mm (older machinery or some brake calipers). And each key from 7/16 in and 12 mm upward is typicaly purchased alone.

Even check the little ones. Keys 0.028 and 0.035 in never come in sets. The metric dimension table shows that the 0.7 and 0.9 mm keys is in instrument territory.

The two normal runs in a single case are combination packs like the 22 piece double sets. There’s nothing more to them, so they gets all the gaps listed above. You should of checked for the missing sizes.