Command A+: benchmark results

Cohere's first Apache 2.0 open frontier MoE (218B total, 25B active), unifying the Command A line. Provider: Cohere. Released 2026-05-20. Access: Open.

Unified ELO 1552 ± 1, rank #426 of 1392 rated models, from 47 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA IFBench73.95Accuracy (%)91.6
LLM Stats (MathVista)80.6Score (%)86.5
CritPt30Accuracy (self-reported)80.8
AA Omniscience-4.02Score78
AA TAU-2 Bench80.7Accuracy (%)71
AA Terminal-Bench Hard25Accuracy (%)66
LLM Stats (CharXiv-D)88Score (%)62.5
LLM Stats (WMT24++)81Score (%)61.5
AA Humanity's Last Exam11.96Accuracy (%)60.2
BenchmarkList ECI118.87Capability Index (ECI)60.2
Artificial Analysis Intelligence Index15.6Intelligence Index59.6
AA GPQA Diamond76.06Accuracy (%)58.8

Interactive version: theaggregate.ai/model?slug=command-a-plus · How It Works · Data refreshed daily, snapshot 2026-09-05.