Command A+
Editorial notes
Lanzado 20 mayo 2026. Nuevo flagship que unifica Command A + Command A Reasoning en un solo modelo. 218B sparse MoE / 25B activos. Corre en 1x B200 o 2x H100 @ W4A4. 48 idiomas, native citation grounding. Apache 2.0, open-weights en HF (CohereLabs/command-a-plus-05-2026). 128K input / 64K max generation. Token pricing no publicado (sales-gated; solo open-weights gratis). Otros scores del blog cohere.com: tau2-Bench Telecom 85%, MMMU-Pro 63%, MathVista 80.6%, CharXiv 52.7%, AA Intelligence Index 37 (self-reported, cross-verificar vs artificialanalysis.ai).
Spec sheet
- Empresa
- Cohere
- Pais
- CA
- Tipo
- multimodal
- Release
- 2026-05
- Context
- 131K tokens
- Slug
- command-a-plus
Benchmarks (2)
Cite this model
BibTeX · APA
BibTeX
@misc{frontier-command-a-plus,
title = {Command A+},
author = {{Cohere}},
year = {2026},
note = {Frontier Benchmarks AI atlas. Accessed 2026-06-17},
url = {https://frontierbenchmarks.com/models/command-a-plus}
} APA
Cohere (2026). Command A+ [Large language model]. Frontier Benchmarks AI. Retrieved 2026-06-17, from https://frontierbenchmarks.com/models/command-a-plus
Citation refleja la pagina del atlas, no el paper original del modelo. Para el paper, ve a la seccion "Recursos" arriba.