Zero Motorcycles VIN decoder

Every road-legal motorcycle sold in the United States since 1981 carries a 17-character vehicle identification number laid out to the same federal rule. 49 CFR 565.13(b) is one sentence long: "Each VIN shall consist of seventeen (17) characters." Section 565.15 then splits them into four sections: (a) positions 1-3, which "shall uniquely identify the manufacturer and type of the motor vehicle"; (b) positions 4-8, which "shall uniquely identify the attributes of the vehicle"; (c) position 9, the check digit; and (d) positions 10-17. Most decoders describe the same layout as three parts — WMI, descriptor and identifier — by folding the check digit in with the descriptor. Either framing gives the same character boundaries.

Zero Motorcycles follows the rule exactly and fills the discretionary positions with its own scheme. The table below is the map. Every row after it in this article expands one line of it.

Two properties of the standard are worth stating before the tables, because they explain things that otherwise look like gaps. First, the letters I, O and Q are not VIN characters at all — they are excluded by ISO 3779 to stop them being confused with 1 and 0, and you can see the consequence directly in the federal year-code table below, which skips them. Second, 49 CFR 565.15(d) requires that for a motorcycle "the last four (4) characters shall be numeric". Zero goes one better and makes the last five numeric: all 23 VINs in our corpus have digits in positions 13 through 17.

The first three characters are the World Manufacturer Identifier, assigned by the national authority — in the United States, SAE on behalf of NHTSA. Zero holds two of them, plus a third identifier that is not a WMI at all, and this is the single most confusing thing about Zero VINs, because a Zero whose number starts PAG looks to most software like a different manufacturer entirely.

Both records are public and can be checked in one request each against NHTSA’s Product Information Catalog and Vehicle Listing (vPIC). Retrieved 29 July 2026, verbatim:

49 CFR 565.15(b) is deliberately loose about these five characters: "The second section shall consist of five characters, which occupy positions four through eight (4-8) in the VIN. This section shall uniquely identify the attributes of the vehicle as specified in Table I." Table I asks a motorcycle manufacturer for "Make, type of motorcycle, line, engine type, and net brake horsepower", and leaves the encoding entirely to them. Zero splits the five characters 1 + 2 + 2.

Position 9 is not information about the motorcycle. It is a modulo-11 checksum over the other sixteen characters, and its only purpose is to catch transcription errors. The algorithm is written out in full in the regulation, so there is no ambiguity about it.

49 CFR 565.15(c), verbatim: "This section shall be the check digit whose purpose is to provide a means for verifying the accuracy of any VIN transcription. After all other characters in VIN have been determined by the manufacturer, the check digit shall be calculated by carrying out the mathematical computation specified in paragraphs (c) (1) through (4) of this section."

One character, and the most frequently misread position on any VIN. The code is federally fixed, not a manufacturer choice: 49 CFR 565.15(d)(1) publishes it as Table VII, quoted here verbatim from the current eCFR text.

One character identifying the assembly plant. Twenty-two of the 23 VINs in our corpus carry C; the single PAG bike carries P.

C is Scotts Valley, CA, USA. Every Zero VIN Breakdown sheet from MY2014 to MY2025 prints exactly that string, the community decoder agrees, and NHTSA reports PlantCity SCOTTS VALLEY with PlantCompanyName "Zero Motorcyles Inc." — the misspelling is in the federal dataset, not a typo here. The company was founded in Santa Cruz and moved a few kilometres up the road to Scotts Valley, which is where 380 El Pueblo Road, the address on its filings, is.

P is the Philippines, and it appears from MY2024. Zero's own MY2025 sheets spell it "Phillipines"; the correct spelling is what the decoder publishes. PAG VINs read position 11 now — the branch used to return before reaching it, so every Gen3 bike reported no plant at all, including the one in our corpus, which is built there.

The A4Z off-road line has a plant set of its own: NHTSA accepts C, T and Z in position 11 for it, against C alone for 538, and the three real A4Z VINs are all Z — Banan District, Chongqing, China. That is why the decoder keeps a separate plant table per WMI rather than sharing one.

Position 12 is where the regulation stops caring. 49 CFR 565.15(d)(3) says positions 12 through 17 "shall represent the number sequentially assigned by the manufacturer in the production process" for a high-volume manufacturer. Zero uses only 13 through 17 as the sequence and spends position 12 on a model letter — which is legitimate, is consistent across every 538 VIN we have, and is invisible to any decoder that assumes the regulation’s literal reading. That includes NHTSA’s own.

This is the position that makes a Zero-specific decoder worth writing. On a 2016 FXS the model-line code at positions 5-6 is X8, which means "FX or FXS" and resolves nothing; position 12 is J, which means FXS.

It is also the position with an end date. From model year 2024 Zero's VIN Breakdown diagrams have no Model row at all — the layout runs the plant character straight into the serial, on the FST street sheet, the FST adventure sheet and the X sheet alike — and the one MY2024 bike in our corpus has a digit there. So every entry in the table below is bounded at MY2023, and a 2024 or later VIN gets no model from position 12 rather than a model letter invented out of a serial digit. That bound is data, not a special case in the code, which is the only reason it is reliable.

Every lookup table on this page except the federal year codes is stored as a list of model-year ranges rather than as a single value per code, and every range is bounded at the years its sources actually cover. That is not tidiness. Zero reuses its codes, and a table that cannot say which years a meaning is good for has to publish it for all of them — including the ones it is wrong for.

The reuse is documented, not suspected. The community decoder keys its tables by model year and records M0 as an MX in 2011 and a 9.8 kWh S in 2016; U2 as an XU-LSM in 2011 and an XU in 2012; M4 as a plain S in 2013 and an 8.5 kWh S/SP in 2014. At position 12 it has C changing meaning four times — X in 2011, XD/X in 2012, XD/X/FX in 2013, XD/X/FX/FXL in 2014 — before Zero's own MY2015 FX manual settles it at FX. Flattening all that into one value per code is what this page used to do.

What shipped is the bound, not yet the second meaning. Ask this decoder about a MY2011 M0 today and it returns nothing at all: no model line, no pack size. Ask the old flat table and it confidently returned S/SR/SP with a 9.8 kWh pack — a 2016 street bike's specification asserted onto a 2011 dirt bike, with is_valid true and no error. Silence is the correct intermediate state, and it is what a bounded range buys you before anyone has reviewed the older reading.

Actually publishing those older meanings is a separate job, deliberately not done in the same pass. A second range changes what this page prints and what the backend writes into a database row, so each one needs its own evidence. Today three codes carry more than one range: the motor codes Z4 and Z6, where the disagreement is not across a decade but between two books Zero shipped in the same year, and the platform character S, which names a body style before MY2011 and a platform after it. Everything else is a single bounded range.

Five digits, assigned in production order. 49 CFR 565.15(d) requires the last four characters to be numeric for a motorcycle; Zero makes all five numeric, and all 23 corpus VINs comply. On the A4Z off-road numbers the serial is six digits, running from position 12, because there is no model designator to make room for.

There is nothing to decode here, but there is something to be careful about. The serial is the only part of a VIN that identifies one specific motorcycle rather than a model, so it is the part that turns a VIN into personal data. Every VIN published on this page and in this project’s documentation has two of its serial characters replaced with XX for that reason, and a masked VIN still decodes correctly because nothing reads positions 13-17.

A practical consequence: NHTSA rejects a masked VIN with error 400, "Invalid Characters Present", and names the offending positions, because X is not a digit. That is a good sanity check that you have masked the number properly before publishing it.

Gen3 — the Cypher III generation, SR/F onwards — broke the position-by-position scheme in two ways at once. Some of those bikes moved to the PAG WMI, and all of them started carrying their identity in a three-character family code at positions 4-6 rather than in the position-12 designator.

The decoder handles the two cases on different paths. A PAG VIN takes an early return: platform is fixed to FST, the pack size comes from the platform character at position 4, the model comes from the family code and the plant from position 11. Model line and motor are left null, because the 538 tables do not apply to those positions. A 538 VIN takes the normal path and only falls back to the family code if the position-12 designator produced nothing.

The first two examples are real motorcycles from this project’s log corpus and the third is a real off-road machine found in the wild, all with two serial characters masked. The tables are generated by running each VIN through the same decoder the widget uses, so what you see here is what the code actually produces, not a hand-written illustration.

Almost every consumer VIN decoder on the web is a wrapper around NHTSA’s vPIC API. It is free, it needs no key, and for Zero Motorcycles it is better than its reputation: it knows the make, the model year, the body class, a series name, a motor rating in kW, a battery size and an electrification level. If you only want to know what year a bike is, vPIC will tell you, and you should use it.

The problems are density and position 12. Here is the complete response profile for the masked 2016 FXS used as the first worked example above, so you can reproduce it exactly:

A VIN is a factory record. It is fixed at build and never updated, which makes it excellent for identity and unreliable for state. Five things in particular are routinely inferred from a Zero VIN and should not be.

The Zero owner community reverse-engineered most of this scheme long before we did, and the best single artefact is vin_decoder.php in the zero-motorcycle-community web-tools repository — a PHP form carrying model-line, motor, platform, plant and designator tables that overlap heavily with ours. It deserves better than the polite nod it used to get here. On the most recent pass through these tables it was the primary source for around ten corrections: the model-year code 8 = 2008, which we simply did not have; the three MY2011 model-line codes X0, X1 and E1; the five Gen1 horsepower bands at positions 7-8; the two MY2011 position-12 designators 1 and 2; and the fact that position 4 encodes a body style rather than a platform before MY2011, which moved a threshold in our code that had matched no source at all. Its year-keyed arrays are also the reason we know how many collisions a flat table was hiding.

It is also wrong in places, and the way it is wrong is instructive. On model-line code X5 it reads FX/FXLP where we read FXP/FXLP — and Zero’s own MY2014 FX owner’s manual prints "X5 = 14 MY FXP/FXLP", so our value stands and the community table is the one to correct. The same manual settles the reverse case in its favour on other codes. Neither source is authoritative on its own; the manual is, when it exists, and it does not exist before MY2014.

Our tables were checked against four further sources: Zero’s own VIN Breakdown sheets, of which we hold thirty-eight across model years 2014 to 2026, and which are the only reason this revision could correct anything; the decoder in this project’s backend, which is what writes the model and pack fields on every uploaded log; the model tags embedded in Gen3 MBB firmware headers, which independently confirm the family codes; and the 23 real VINs described above, including one whose battery telemetry decided a disagreement between two documents. Where the community tables and a real log disagree, the log wins; where a manual and a log disagree, we say so rather than choosing.

This page renders the lookup tables directly from the same TypeScript module the widget calls, which is itself generated from the Python decoder and checked in continuous integration — editing either side alone fails the build with a diff. There is no second copy to fall out of date, and the counts in the prose above are interpolated from those same tables rather than typed, because for a while the prose was the second copy. If you find a code we have wrong, or a VIN that decodes to the wrong bike, tell us — a masked VIN and the correct model is enough to fix it.