Delivers advanced reasoning with a minimal compute footprint, multimodal, multilingual across 48 languages, and ensures full data sovereignty for governments and regulated industries worldwide

TORONTO: Cohere, the trusted sovereign AI platform for governments and regulated industries worldwide, today announced the release of Command A+, a powerful open-source Mixture-of-Experts (MoE) model released under an Apache 2.0 license. Purpose‑built for sovereign critical infrastructure, Command A+ gives enterprises and public sector organizations the transparency, control, and efficiency required to deploy AI at scale while upholding the highest standards of security and data sovereignty.

As open‑source AI development becomes increasingly concentrated in a small number of jurisdictions, namely China, organizations operating critical systems face growing concerns around transparency, security, vendor dependence, and long‑term sovereignty. Command A+ addresses these challenges with a fully open‑weight model that can be deployed wherever sensitive data resides-whether in a VPC, on‑premises, or fully air‑gapped-while providing complete operational control and confidence.

“The world is building its future on AI systems that offer no guarantee of control,” said Nick Frosst, Co‑founder of Cohere. “And when only a few big companies control the technology, people have little ability to shape how their AI systems work or evolve. That’s why it’s so important to prioritize sovereignty now and make sure organizations can own their intelligence layer instead of renting it. Command A+ is part of our effort to make that possible. We’re delivering a sovereign, open‑source model built for critical infrastructure that gives people, enterprises, and governments the trust, performance, and efficiency they need to run real‑world systems at scale.”

Built for Enterprise Reality

Command A+ is engineered for the realities of enterprise deployment, where reliability, cost efficiency, and operational performance determine whether a system can be deployed at all. In a world of global compute scarcity, where every token has an opportunity cost, Command A+ delivers advanced reasoning with a minimal compute footprint. Its MoE architecture spans 218 billion parameters while activating only 25 billion per prompt, enabling high‑performance inference on as few as two H100s or a single B200 GPU. This efficiency makes Command A+ practical for private deployments where hardware is fixed and cost predictability is essential.

While many large models are optimized for standardized benchmarks, Command A+ is built for the actual workloads enterprises run under tight compute budgets: complex RAG pipelines, multi‑step SQL generation, and high‑stakes financial document analysis. Its multimodal capabilities extend this performance to visual data, enabling organizations to process charts, PDFs, slides, and complex reports. It balances latency, reasoning depth, tool use, and cost to achieve the equilibrium that makes AI deployable, not just demonstrable.

Expanding Global Access

Command A+ delivers strong multilingual performance for global teams, supporting 48 languages, including all official EU languages. With the EU AI Act now in effect, organizations operating across Europe need models that provide accurate linguistic and cultural understanding. The model also offers robust support for Japanese, Arabic, Hindi, and other widely used global languages, enabling multinational teams to work across regions while meeting local data‑residency and compliance requirements.

Sovereign by Design

Released with fully open weights under Apache 2.0, Command A+ gives organizations full transparency into model behavior and complete control over deployment.

Key sovereignty features include:

• Zero hidden backdoors: Full visibility into model architecture and behavior
• Total data sovereignty: On-premises and private cloud deployment with no external data transmission
• Regulatory alignment: Control and deployment model designed to enable alignment with evolving global compliance and AI governance requirements
• No vendor lock-in: Predictable costs with no licensing restrictions or forced ecosystem dependencies

Across regulated industries, organizations are moving from experimental AI to systems that deliver reliability, security, efficiency, and full operational control like the Cohere North platform. As demand for sovereign AI accelerates, enterprises and public‑sector agencies are prioritizing deployments that give them complete ownership of their data, infrastructure, and model behavior. Command A+ is built to meet these requirements: a transparent, controllable, open‑source model designed for large‑scale production environments.

Find more technical information in our blog.

About Cohere

Cohere, founded 2019, is a security-first enterprise AI leader building foundation models and end-to-end products to solve real-world business problems. We partner with organizations to deliver seamless integration, customization, and user-friendly solutions. Our all-in-one platform provides maximum security, privacy, and deployment flexibility across clouds, private environments, and on-premises. Headquartered globally in Toronto and San Francisco, Cohere operates additional offices in London, New York, Montreal, Paris, and Seoul, serving customers worldwide.

The company has raised ~$1.6BUSD from strategic tech investors (Nvidia, AMD Ventures, Salesforce Ventures, Oracle, Cisco), institutional investors (Radical Ventures, Inovia Capital, PSP Investments, HOOPP, BDC, Nexxus), and AI pioneers including Geoffrey Hinton, Fei-Fei Li, Pieter Abbeel, and Raquel Urtasun. For more information, visit cohere.com.

Fonte: Business Wire