VIN Decoder API: bulk NHTSA lookup, specs and recalls

Decode one VIN or 100,000 in a single run with the US government's official vehicle data (NHTSA vPIC): make, model, model year, trim, body, engine, drive, transmission, GVWR and assembly plant, one clean row per VIN, ready for Excel, CSV, JSON or your app. Every VIN is also checked against the federal VIN rule (49 CFR 565): check digit, model year code, maker code, plant and serial number, with plain-English warnings when something is wrong. Model year 2026 VINs from 19 makers are also read against the maker's own VIN code tables filed with NHTSA. Recalls, owner complaints and NHTSA's 5-Star crash test ratings can be added for each vehicle.

No API key, no sign-up anywhere else, no subscription: USD 1.00 per 1,000 decoded VINs.

Who it's for

• Car dealers and auction sites: check and complete a whole inventory or a lot list from its VINs in seconds.
• Fleet managers: one sheet with every vehicle's specs, plant and open recall campaigns for its model year.
• Insurers, lenders and leasing: catch mistyped or fake VINs (check digit, model year, test patterns) before they reach a quote or a contract.
• Parts shops and marketplaces: turn a VIN into make, model, year, engine and drive to match parts.
• Developers: a VIN decoder API you call from code, with the same columns on every row.

What you get for each VIN

Group	Fields
Vehicle (NHTSA vPIC)	make, model, modelYear, trim, series, bodyClass, vehicleType, doors, seats, driveType, engine (one line), engineCylinders, displacementL, engineHP, engineModel, fuelTypePrimary, electrificationLevel, batteryKWh, transmissionStyle, transmissionSpeeds, gvwr, curbWeightLB, manufacturer, plantCity, plantState, plantCountry
VIN check (offline, 49 CFR 565)	valid, checkDigit, checkDigitExpected, checkDigitValid, modelYearCode, modelYearCandidates (both years a code can mean), modelYearOffline, wmi, manufacturerId, vds, plantCode, serialNumber
Flags	flags (codes) and warnings (plain English): wrong length, illegal characters, I/O/Q typos with a corrected VIN, check digit mismatch, invalid year code, model year after next year, test patterns, NHTSA's own error codes explained
Maker's VIN filing (model year 2026, Ford 2027)	makerFilingModel, makerFilingDecode (each coded position as the maker wrote it: line, series, body, engine, restraints, GVWR), makerFilingPlant, makerFilingSource (file and date), makerFilingLink (the PDF at NHTSA)
Recalls (option)	recallCount, recallCampaigns, recallParkIt, recallParkOutside, recallModels (the NHTSA name used)
Complaints (option)	complaintCount, complaintCrashes, complaintFires, complaintInjuries, complaintDeaths, complaintTopComponents
Crash test ratings (option)	ratingOverall, ratingFrontCrash, ratingSideCrash, ratingSidePole, ratingRollover (NHTSA 5-Star stars), rolloverRiskPercent, ratingsVehicle (the rated version used)
Nested details (option)	nhtsaFields (every non-empty NHTSA field for the VIN), makerFiling (position by position), recalls (subject, summary, consequence, remedy, report date, units affected)

Every row also says where each part came from (source) and its status:

status	Meaning	Billed as
decoded	NHTSA decoded the VIN cleanly (make and model)	VIN decoded
decoded_with_warnings	NHTSA decoded it but reported issues (see flags)	VIN decoded
partial	NHTSA returned the make but no model (common for VINs of cars not sold in the US)	VIN checked
not_decoded	NHTSA returned no make (for example an unregistered maker)	VIN checked
invalid	not a valid 17-character VIN; not sent to NHTSA	VIN checked
nhtsa_unavailable	NHTSA did not answer after retries; the offline check is still in the row	VIN checked
offline	you switched NHTSA off: offline check and maker filing only	VIN checked

Why this decoder

• Bulk and fast: 50 VINs per NHTSA request (the most NHTSA allows), about 1,000 VINs in 30 seconds. Paste a list or give the link to a CSV (a Google Sheet published as CSV works); the VIN column is found for you. Duplicates are decoded once, and text that is not a VIN (makes, prices, stock numbers) is skipped, not billed.
• Explains the VIN, not just looks it up. The check digit with the value it should have, both model years a code can mean and which one applies, the I/O/Q typo fixed when the check digit proves it. Works even when NHTSA is down.
• No false alarms on non-US VINs. NHTSA marks many European and Asian home-market VINs as "check digit does not calculate". The check digit is only required for vehicles made for the US and Canada; the warning says so.
• The maker's own codes for model year 2026 (Ford 2027): Ford, GM (Chevrolet, GMC, Buick, Cadillac), Stellantis (Jeep, Ram, Dodge, Chrysler), Toyota and Lexus, Honda and Acura, Hyundai and Genesis, Kia, Nissan and Infiniti, Subaru, Mazda, Mitsubishi, Tesla, BMW and MINI, Mercedes-Benz, Volkswagen, Audi, Porsche, Volvo and Polestar, Jaguar Land Rover.
• Recalls you can trust. NHTSA's safety data names vehicles differently from its VIN decoder (the decoder says "Silverado", the recall data "SILVERADO 1500"; 2026 models are listed as "TAHOE ICE", "MODEL Y BEV"). This decoder matches the names, including the powertrain, and never borrows another version's recalls (a Q5 SUV does not get the Q5 Sportback's). When NHTSA does not find the vehicle, the row says not found, never "0 recalls".
• NHTSA trouble never fails your run. Retries with backoff; a VIN NHTSA cannot answer still gets its row and a clear reason; you are billed the lower price for it.

Input

{
  "vins": ["1HGCM82633A100000", "1GCUYDED0LZ100001, 2020", "1FTMW1KC9VFA00001"],
  "recalls": true,
  "complaints": false
}

Field	Default	Notes
vins		One VIN per line (the form starts with 4 examples). Spaces, dashes and lower case are fine. Add the model year after a comma if you know it. Up to 100,000 per run
vinListUrl		A .txt or .csv file (up to 5 MB). The VIN column is found by its header or content; a Year or Model year column is used. Web pages and Excel files are refused (save as CSV)
nhtsa	true	Off = offline check only (cheaper)
recalls	false	Recall campaigns for each vehicle's make, model and model year
complaints	false	Owner complaint summary for the make, model and model year
ratings	false	NHTSA 5-Star crash test ratings for the matching rated version
details	false	Nested blocks: all NHTSA fields, the maker's filing position by position, each recall's text
fixIoq	true	Read I as 1 and O, Q as 0 when the check digit then holds (the row says so)
dedupe	true	Same VIN and year given twice = one row

Output example (one row, shortened)

{
  "vin": "1GCUYDED0LZ100001",
  "status": "decoded",
  "make": "CHEVROLET",
  "model": "Silverado",
  "modelYear": 2020,
  "trim": "LT",
  "series": "1500",
  "bodyClass": "Pickup",
  "driveType": "4WD/4-Wheel Drive/4x4",
  "engine": "5.3 L, 8 cyl, L84 - DI, DFM, ALUM, GEN 5, VAR 2",
  "transmissionStyle": "Automatic",
  "transmissionSpeeds": 8,
  "plantCity": "FORT WAYNE",
  "plantCountry": "UNITED STATES (USA)",
  "checkDigitValid": true,
  "modelYearCandidates": "2020 or 1990",
  "flags": null,
  "recallsStatus": "ok",
  "recallModels": "SILVERADO 1500",
  "recallCount": 12,
  "source": "49 CFR 565 (offline) + NHTSA vPIC + NHTSA recalls"
}

A 2026 vehicle also gets the maker's filing, for example (Ford's model year 2027 filing):

{
  "makerFilingModel": "Ford F-150",
  "makerFilingDecode": "4 (M): Hydraulic brakes; GVWR E (6001 - 7000 lb); restraints: Manual Belts, Driver Frontal, Pass. Frontal, Side inflatable 1st Row; 5-7 (W1K): F-Series, F150 XL, 4x2, SuperCrew; 8 (C): 3.0L GTDI, V-6, Gasoline, Ford",
  "makerFilingPlant": "Dearborn: Dearborn, Michigan",
  "makerFilingSource": "Ford Motor Company, model year 2027 VIN filing to NHTSA (49 CFR 565.16): 2027 MY VIN Attachment -Rev 3.pdf, August 6, 2026"
}

The example VINs in this page and in the input are made up on real VIN patterns, with a correct check digit.

The dataset has five views: Overview, Specs, VIN checks, Maker's VIN filing, Recalls and complaints. The run summary (record OUTPUT) counts rows per status, duplicates dropped and NHTSA requests, and says why a run stopped early if it did.

Pricing

Pay per event, no subscription:

Event	Price
VIN decoded (NHTSA returned the vehicle)	USD 1.00 per 1,000
VIN checked (invalid, offline mode, or NHTSA could not answer)	USD 0.30 per 1,000
Recalls added to a VIN	USD 1.00 per 1,000
Complaints added to a VIN	USD 1.50 per 1,000
Crash ratings added to a VIN	USD 1.00 per 1,000
Run start	USD 0.002

Recalls, complaints and crash ratings are billed only on rows where NHTSA answered (status ok). Set a maximum cost per run and the decoder stops cleanly before it, keeping every row decoded so far.

Good to know

• Recalls and complaints take longer: NHTSA's safety service is asked politely (about 3 requests a second), so a list of all-different vehicles runs at roughly 1 to 2 seconds per vehicle; a fleet of the same models is much faster (each make, model and year is looked up once per run).
• Recalls are per model year, not per VIN. NHTSA's public data lists the recalls for a make, model and model year; it does not say whether this exact vehicle is included or already repaired. Check a single VIN at nhtsa.gov/recalls or with the maker.
• Before 1981 VINs were not standardized (often fewer than 17 characters). No government source decodes them; such VINs are flagged, not guessed.
• NHTSA covers vehicles sold in the US. VINs of vehicles never sold there often decode only partly (make, plant country); missing fields stay empty. Nothing is filled in by guesswork.
• Model year 2026 maker filings come from NHTSA's Manufacturer Information Database (the makers' own 49 CFR 565.16 submissions). Other years use NHTSA's decode only.
• Data sources: NHTSA vPIC (vpic.nhtsa.dot.gov/api), NHTSA recalls and complaints (api.nhtsa.gov), NHTSA Manufacturer Information Database. This is not a vehicle history report: no owners, titles, accidents or mileage.

FAQ

Is this a free VIN decoder? The data is NHTSA's free public data; the tool is pay per use (USD 1.00 per 1,000 decoded VINs, nothing per month). Apify's free plan credit covers thousands of VINs.

How many VINs can I decode at once? Up to 100,000 per run, 50 per NHTSA request; 1,000 VINs take about 25 seconds.

Can I upload a spreadsheet? Yes: paste the VIN column, or give the link to a CSV file (a Google Sheet published as CSV works). The VIN column and a Year column are found for you; other columns are ignored.

Does it give vehicle history (accidents, owners, mileage, title)? No. That data is not public. This is a VIN decoder: what the vehicle is, as its maker reported it to NHTSA, plus NHTSA's recalls and complaints for its model.

Which crash test rating does a VIN get? NHTSA rates versions of a model (for example a 2020 Silverado 1500 crew cab 4WD, extended cab 2WD...). The decoder picks the version that matches the VIN's drive, cab and doors. When several fit and their stars differ, all are listed in ratingsCandidates and the row is not billed for ratings. Models NHTSA has not tested say not_rated, never zero stars.

Does it tell me if a recall was repaired on this car? No: NHTSA's public data lists recalls by make, model and model year. Check one VIN at nhtsa.gov/recalls or with the maker.

Does it work for cars outside the US? The offline check works on any 17-character VIN. NHTSA's decode covers vehicles sold in the US; others usually decode only partly (make, plant country) and are billed at the lower check price.

What happens to VINs from before 1981? They were not standardized (often fewer than 17 characters) and no government source decodes them, so they are flagged invalid, never guessed.

What if NHTSA is down? The tool retries, then saves each VIN with its offline check and a clear reason, billed at the lower check price. Run those VINs again later.

Use it from code

curl -X POST "https://api.apify.com/v2/acts/usefulapi~vin-decoder/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"vins": ["1HGCM82633A100000", "1GCUYDED0LZ100001"], "recalls": true}'

Questions or a VIN that decodes wrong? Open an issue on the Issues tab with the VIN and what you expected.