Command-R+ (Apr '24): benchmark results
Cohere Labs Command R+ April 2024 checkpoint for multilingual grounded generation and multi-step tool use. The publisher specifies 104B parameters and a 128K context; this row preserves the April release separately from the August revision. Provider: Cohere. Released 2024-04-04. Access: Open.
Aggregate Rating 1445 ± 28, rank #3893 of 10870 rated models, from 18 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Enkrypt AI - Jailbreak Risk | 9.28 | Risk Score | 50 |
| PLCC - Culture & Tradition | 52 | Accuracy (%) | 29.1 |
| PLCC - Vocabulary | 46 | Accuracy (%) | 28.9 |
| PLCC - Art & Entertainment | 39 | Accuracy (%) | 23 |
| PLCC - Overall | 49.33 | Mean category accuracy (%) | 22.4 |
| AA Humanity's Last Exam | 4.63 | Accuracy (%) | 22 |
| PLCC - History | 61 | Accuracy (%) | 22 |
| PLCC - Geography | 53 | Accuracy (%) | 18.7 |
| PLCC - Grammar | 45 | Accuracy (%) | 15.4 |
| Enkrypt AI - Toxicity Risk | 12.27 | Risk Score | 12.2 |
| AA LiveCodeBench | 12.17 | Pass@1 (%) | 11.4 |
| AA MMLU-Pro | 43.18 | Accuracy (%) | 10.1 |
Interactive version: theaggregate.ai/model?slug=command-r-plus-apr-24 · How It Works · Data refreshed daily, snapshot 2026-11-12.