2026 Car Safety Report

What the data shows

Also new: The July 2026 driver death-rate release covers newer used-car model years and risk to other drivers.

This report covers 198 current IIHS vehicle records: 179 from model year 2026 and 19 from model year 2027. Of these, 136 have all results required for our ranking; 62 remain unranked because a required result is missing.

“2026” is the award and report year, not a claim that every vehicle was crash-tested this year. IIHS can carry a test result across model years where it applies. Awards, scores and coverage below reflect our saved 2026-09-16 dataset; later IIHS announcements may change the totals.

Rear restraints and dummy motion are rated below Good in 105 records, and not reported in 44. This is one subrating within the updated moderate overlap test, not a separate overall rear-seat rating.

Use the 2026 car safety ratings lookup to compare individual models, or the full ranking to explore both the standard and rear-seat-priority scores.

Ten highest-scoring models in the snapshot

These are the first ten records in the current ranking. Detailed injury ratings break equal overall scores; records matching on both share a rank. A tie may extend beyond this selection. Body styles and powertrains have separate records where IIHS lists them separately.

The score summarizes the available evidence; it does not predict an individual model’s fatality rate. Check the exact trim and build-date restrictions on the linked model and IIHS pages before choosing a vehicle.

Brand scorecard: weighted by sales volume

Current ratings, weighted by January–June 2026 U.S. sales. The mean is the sum of each current score × its model’s sales, divided by sales with complete, compatible scores. High-volume models therefore count more. This changes brand aggregation only; individual vehicle scores and ranks are unchanged.

A range shows the possible mix of separately rated variants when their sales split is unpublished. We count a combined sales total once. Unmatched models, unresolved body styles and incomplete scores are excluded from the mean and reduce its sales coverage. Do not compare a brand covering a small slice of sales with one covering nearly its whole lineup as though the evidence were equal.

The TSP+ sales-mix proxy divides sales mapped to current TSP+ records by all reported brand sales. A range reflects award differences within mapped variants. Unmatched sales add nothing to this proxy; they are not rated failures. Current model years, build dates and required equipment still apply: this is a sales-mix comparison of current designs, not the percentage of vehicles delivered in the first half that actually qualified for an award. It is not a death-risk estimate.

Brands are alphabetical because sales coverage and variant uncertainty differ. Figures labeled “Compiled quarterly figures” use GCBC’s Q1 and Q2 model tables and brand totals; they have not all been independently checked against manufacturer releases. Manufacturer tables replace those figures where checked. Estimates without verified U.S. model splits are withheld. Each brand page shows the complete mapping and exclusions.

Sources and limits

Ratings and awards come from the IIHS award index and the model pages linked above. Published class-level driver fatality rates provide a small contextual input, as documented in our methodology. NHTSA stars are displayed separately on model pages.

This report does not measure year-over-year brand improvement: the tested model mix and protocols differ across years. It does not fill missing tests with assumed passes, convert foreign-market results to US ratings, or claim that second-row tests measure third-row protection.