
By Kasia Olejarczyk · 28 July 2026
Almost every torque wrench on the market is built to one international standard, and almost nobody buying one has read it. That is a shame, because the standard already answers the questions buyers ask.
Which type, and which class. How far the tool is allowed to be wrong, and when it has to go back for checking. ISO 6789 sets all four.
This guide walks that standard back into plain language, in the order the decision actually gets made. Read it once and a spec sheet stops being marketing copy.
Everything below comes from the standard itself.
The number that matters in a bolted joint is clamp load, and no torque wrench measures it directly. Torque is the cheap, practical stand-in for it.
Because it is only a stand-in, the whole chain has to be controlled.
Without a torque wrench, guessing fails in both directions. Under-tighten and the joint works loose under load, and over-tighten and you stretch the fastener past its useful range.
ISO 6789-1 divides every hand torque tool into two types. The split is about what the tool does when you reach the target value.
A Type I tool indicates the torque being applied, through a mechanical scale, a dial or an electronic display. You read the value as it climbs, and you decide when to stop. Nothing in the tool intervenes.
A Type II tool senses the applied torque by comparing it with a self-contained standard inside the tool. The familiar click wrench sits here, and so does every preset tool.
Under each type the standard runs a set of class letters, and they describe the mechanism rather than the brand name.
| Standard label | What it means | Common name |
|---|---|---|
| Type I, Class A | Torsion or flexion bar, reads on a pointer | Beam torque wrench |
| Type I, Class B | Rigid housing with scale or dial | Dial torque wrench |
| Type I, Class C | Rigid housing, electronic measurement | Digital torque wrench |
| Type II, Class A | Adjustable, graduated or with display | Click torque wrench |
| Type II, Class B | Fixed adjustment, not user set | Preset torque wrench |
The standard is blunt about the word everyone uses. It notes that "accuracy" is still used in place of maximum permissible deviation, and that this is not technically correct. Accuracy is not a quantity, so it carries no numerical value.
What a torque wrench datasheet really quotes is the maximum permissible relative deviation.
Below that figure the standard is more forgiving, because small absolute errors weigh heavily in percentage terms.
Above it, the better classes have to hold a tighter band. That single line explains most of the price difference between two torque wrenches that look identical.
| Type and class | Up to 10 N·m | Above 10 N·m |
|---|---|---|
| Type I, Classes A and D | ±6 % | ±6 % |
| Type I, Classes B, C and E | ±6 % | ±4 % |
| Type II, Classes A, B and C | ±6 % | ±4 % |
| Type II, Classes D, E, F and G | ±6 % | ±6 % |
A maker may claim a tighter figure than the table, and every test result then has to fall inside that claim. A tool that works in both directions has to meet its limit in each direction.
The standard ties the maximum torque of a torque wrench to its output drive size. A bigger job therefore needs a bigger drive, not a longer pull on the same one.
| Square drive | Common name | Maximum torque |
|---|---|---|
| 6,3 mm | 1/4 inch | 30 N·m |
| 10 mm | 3/8 inch | 135 N·m |
| 12,5 mm | 1/2 inch | 340 N·m |
| 20 mm | 3/4 inch | 1 000 N·m |
| 25 mm | 1 inch | 2 100 N·m |
A graduated tool is specified from its lowest marked value up to its full maximum, so buy for the band you work in rather than the top number. Scale increments are capped at 5 % of that maximum, which tells you how finely you can really set it.
Conformance testing runs every torque wrench through 5 000 cycles in each working direction, at a rate between 5 and 20 cycles a minute. It must still hold its deviation limit afterwards. Tools also face an overload test at 125 % of maximum torque, three times in each direction.
That tells you two useful things at once. The tool is built to be worked hard, and the mechanism is expected to drift as it goes.
Part 1 of the standard covers design and manufacture, and it deliberately dropped the old requirement for a calibration certificate. Traceable calibration now lives in Part 2, which also sets out how uncertainty is calculated.
Makers have to test how changing geometry affects torque output, including flexible head ratchets and extension bars meant to reduce operator effort.
For general controlled tightening, a Type II adjustable graduated wrench, which is the everyday click type. Choose an indicating Type I tool when you need to watch the value climb rather than be stopped at it.
On deviation, usually yes, by the standard's own tables. A click wrench sits in Type II Class A at ±4 % above 10 N·m, while a beam wrench sits in Type I Class A at ±6 % throughout.
It is the furthest the tool may sit from the value you set, measured against that target. On a 200 N·m setting a conforming tool may land anywhere between 192 and 208 N·m. Whether that matters is a question about the joint, not about the tool.
Follow the maker's instruction sheet, because designs differ on this. What the standard does establish is that the mechanism is a wearing item, tested over thousands of cycles and expected to drift.
There is no single interval written into the standard. Set one from how hard the tool works and how critical the joints are, and use a traceable calibration where the joint really matters.
Only if you know what it does to the reading. The maker has to test and publish the effect of flexible heads and extension bars, so the answer is in the paperwork rather than in the workshop.
Start with the type, because indicating and setting tools solve different problems. Then read the class letter, which tells you both the mechanism and the deviation band you are entitled to. After that the choice narrows quickly.
Drive size caps your torque, and the marked range decides where the tool is honest.
The calibration paperwork decides whether anyone else will accept your number. A torque wrench chosen that way costs no more than one chosen off a picture, and it will still be right in three years.
Related categories on EYBY: