Llama-Primus-Merged: benchmark results

Provider: Meta.

Aggregate Rating 1490 ± 49, rank #2461 of 10556 rated models, from 9 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AthenaBench - ATE33.8Accuracy (%) extracting ATT&CK techniques from threat report75
AthenaBench43.7Combined score (%): mean of the six task scores, full benchm57.1
AthenaBench - RCM56.6Accuracy (%) mapping vulnerability descriptions to their CWE44.4
AthenaBench - CKT76.3Accuracy (%) on the CTI knowledge test (3,000 questions)42.9
AthenaBench - RMS6.6F1 (%) of proposed ATT&CK mitigations against the reference 37.5
AthenaBench - TAA17Accuracy (%) attributing anonymised incident reports to a th31.2
AthenaBench - VSP71.9Accuracy (%) = 1 - MAD/7.7 of predicted CVSS base scores ove22.2
KaliBench16.3Exact Correct (%), Unrestricted mode: share of the 5,000 ver17.4
RedSage-Bench74.81Macro Accuracy (imported)6.2

Interactive version: theaggregate.ai/model?slug=llama-primus-merged · How It Works · Data refreshed daily, snapshot 2026-11-17.