# CVSS v4.0 changes the metrics, the scoring method and the labeling, so v3.1 and v4.0 scores must not share one sorted list

COMPARISON, read 10 Nov 2026

**The short answer.** CVSS v4.0, published 2023-11-01, adds Attack Requirements, a Threat group and a Supplemental group, scores by 270 MacroVectors instead of the v3.1 formula, and requires labels such as CVSS-BT. In the KEV Lag Ledger, 313 of 318 scores are v3.1, 5 are v3.0 and none is v4.0.

- **Maintained by:** FIRST CVSS-SIG, both versions
- **Scale:** 0.0 to 10.0 in both, with the same qualitative bands
- **Current version:** v3.1 (archive) and v4.0 (specification document v1.2, 2024-06-18)
- **Updated:** v4.0 specification 1.0 on 2023-11-01, 1.1 on 2023-11-09, 1.2 on 2024-06-18
- **Answers:** Both: the severity of a flaw; v4.0 adds a Threat group for exploit maturity

## Side by side

Both specifications are published by FIRST and share the 0.0 to 10.0 scale and the qualitative bands: None 0.0, Low 0.1 to 3.9, Medium 4.0 to 6.9, High 7.0 to 8.9, Critical 9.0 to 10.0. They differ in groups, metrics, method and labeling. The table lists the differences that matter to a program.

| Property | CVSS v3.1 | CVSS v4.0 |
| --- | --- | --- |
| Metric groups | Base, Temporal, Environmental | Base, Threat, Environmental, Supplemental |
| Impact metrics | C, I, A with Scope (Unchanged or Changed) | VC, VI, VA for the vulnerable system and SC, SI, SA for subsequent systems |
| Attack conditions | Attack Complexity | Attack Complexity and Attack Requirements (AT) |
| User Interaction | None or Required | None, Passive or Active |
| Exploitation input | Temporal: Exploit Code Maturity | Threat: Exploit Maturity (E) |
| Scoring method | Closed formula with Roundup | 270 MacroVectors, expert ordering, interpolation |
| Labeling | Not required | CVSS-B, CVSS-BT, CVSS-BE, CVSS-BTE |

## Scoring method

The v3.1 formula can be recomputed by hand: ISS = 1 - (1 - C)(1 - I)(1 - A), Impact = 6.42 x ISS (unchanged scope), Exploitability = 8.22 x AV x AC x PR x UI, and BaseScore = Roundup(Min[Impact + Exploitability, 10]). The worked case AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H gives 9.8.

The v4.0 specification gathers 15 million CVSS-BTE vectors into 270 equivalence sets, uses expert comparison data to order them from least to most severe, and fixes the boundaries between qualitative bands to match the v3.x boundaries. A score is a lookup plus interpolation. An auditor needs the FIRST calculator or a verified implementation to reproduce it.

The metric lists are worth stating in full. Under v4.0 the Base group has AV, AC, AT, PR, UI and the six impact metrics VC, VI, VA, SC, SI, SA. The Supplemental group contains Safety, Automatable, Recovery, Value Density, Vulnerability Response Effort and Provider Urgency, which the specification says consumers may weigh more, less or not at all. Under v3.1 the Base group has AV, AC, PR, UI, S, C, I and A.

## Exploitation and labeling

v3.1 has a Temporal group that scales the Base Score down: TemporalScore = Roundup(BaseScore x ExploitCodeMaturity x RemediationLevel x ReportConfidence). v4.0 replaces it with a Threat group of one metric, Exploit Maturity, and states that when Threat and Environmental metrics are not defined the score assumes the highest severity.

The labeling rule is the largest practical change. The v4.0 specification states that numerical scores should carry nomenclature because the usefulness of a score for prioritization is proportional to the metrics used to produce it. A column of v4.0 scores without labels mixes CVSS-B and CVSS-BT and cannot be sorted honestly.

A reading of the vector strings makes the change visible. A v3.1 vector begins CVSS:3.1/ and lists AV, AC, PR, UI, S, C, I, A. A v4.0 vector begins CVSS:4.0/ and lists AV, AC, AT, PR, UI, VC, VI, VA, SC, SI, SA. A parser that assumes the v3.1 field list will fail or misread v4.0, so a data pipeline should branch on the prefix.

## In the data

The KEV Lag Ledger takes its severity from the NVD record, v3.1 first, then v3.0, v4.0 and v2. All 318 additions were scored: 313 with v3.1 and 5 with v3.0, and none fell through to v4.0. For this set of exploited CVEs, the v3 family was what the NVD and CNAs had published.

The NVD news page shows v4.0 data quality: on 2026-04-28 NIST announced that about 4,500 CVE records, 19 percent of those with a v4.0 score, carried a numerical score higher than the correct value, and fewer than 30 carried a lower one. NIST said it would correct them by script on 2026-04-28. A program that imported v4.0 scores before then should confirm that it picked up the correction.

Who scores matters as much as which version. The ledger's score column names its source: of the 318 rows, 130 come from nvd@nist.gov and the rest from CNAs. Since 2026-04-15 the NVD does not routinely add a score where a CNA has supplied one, so whether a CVE has a v4.0 score depends on whether its CNA chose to publish one.

## Migration practice

Run both versions in parallel for a period. Store each vector string with its version, compare the band each method assigns to the same CVEs, and look at the CVEs where the bands differ. Do not assume a mapping from a v3.1 score to a v4.0 score.

Change the policy text at the same time: name the version in every threshold, such as CVSS-BT 4.0 of 7.0 or above, and set a date after which v3.1 scores are shown for reference only.

Set a cutover plan with dates, and name an owner for each phase. Phase one stores both versions and reports v3.1. Phase two reports v4.0 where present and falls back to v3.1 with a flag. Phase three retires v3.1 from thresholds once the share of open findings with a v4.0 vector is high enough to be useful, a figure each program has to measure for itself.

## Which to cite

Where a CVE carries only v3.1, cite v3.1. Where it carries both, cite the version your policy names. Where the policy names v4.0 and a CVE has only v3.1, record that the CVE was scored under the older method and apply the same band cutoffs, noting the substitution.

Neither version replaces the exploitation signals. The ledger's 10.4 percent of KEV additions under CVSS 7.0 is a v3 figure, and the same test on v4.0 would need a fresh run.

## Verdict by situation

- **Federal agencies and contractors:** Rank by KEV status first and cite CVSS only with its version and vector string, since BOD 26-04 uses technical impact rather than either CVSS version.
- **Large enterprises with ten or more scanners:** Rank by exploitation evidence first and cite CVSS-BT 4.0 where the source scored it, with v3.1 retained for reference during the migration.
- **Managed security providers:** Rank by exploitation evidence first, and report both CVSS versions side by side because client exports may carry either one.
- **Mid-size teams replacing Kenna or a spreadsheet:** Rank by EPSS and KEV first and keep to one CVSS version, whichever the new platform shows, so the severity column does not mix methods.
- **Application security and infrastructure in one queue:** Rank by CVSS v4.0 first for application findings that testing tools score with Threat and Supplemental metrics, and by v3.1 for older infrastructure CVEs that carry only the older score.

## Other comparisons

- [EPSS vs CVSS](/compare/epss-vs-cvss/)
- [CVSS vs KEV](/compare/cvss-vs-kev/)
- [EPSS vs KEV](/compare/epss-vs-kev/)
- [SSVC vs CVSS](/compare/ssvc-vs-cvss/)
- [SSVC vs EPSS](/compare/ssvc-vs-epss/)
- [KEV vs vendor intelligence](/compare/kev-vs-vendor-intelligence/)
- [CVSS vs EPSS vs KEV](/compare/cvss-vs-epss-vs-kev/)

## Sources

- FIRST, CVSS v3.1 specification: https://www.first.org/cvss/v3.1/specification-document (read 10 Nov 2026)
- FIRST, CVSS v4.0 specification (version history, MacroVectors, nomenclature): https://www.first.org/cvss/v4-0/specification-document (read 10 Nov 2026)
- NVD news (2026-04-28 v4.0 score correction): https://nvd.nist.gov/general/news (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://epssvscvss.com/compare/cvss-v3-1-vs-v4-0/
