Cohere released Command A+ as open weights. What should its open-weights strategy be, and how does it make money from it?

Introduction

For most of its history, Cohere released model weights for research under non-commercial licences (for example through Cohere Labs' Aya and Command R releases) while selling commercial access through its API and private deployments. In May 2026, Cohere announced Command A+, an enterprise model released under the permissive Apache 2.0 licence, meaning anyone can run, modify and use it commercially. That is a significant strategic shift for a company whose revenue depends on selling model access.

The prompt: "Cohere has released Command A+ as open weights. What should Cohere's open-weights strategy be over the next two years, and how does it make money from it?"

Key business objectives:

1. Grow enterprise revenue, mostly from private deployments and North.
2. Win trust with governments and regulated industries that want inspectable, controllable models.
3. Expand developer mindshare against Meta's Llama, Mistral AI, Alibaba's Qwen and other open models.
4. Avoid cannibalising paid model revenue.

I'll clarify goals, analyse the market, define a vision and roadmap, prioritise, plan execution, set KPIs, and manage risks.

Step 1

Clarify the Strategic Goals (3-4 minutes)

• What share of Cohere's revenue comes from model access versus platform and services?

Why it matters: If model access dominates, open weights risk cannibalisation. Expected answer: Reports suggest most revenue comes from private deployments, where customers pay for the full stack, support and North rather than per-token API calls. Impact on approach: Open weights can be a funnel into paid platform deals rather than a threat.

• Who is the open release meant to attract?

Why it matters: Developers, governments and partners need different packaging. Expected answer: Sovereign and regulated buyers first, developers second. Impact on approach: Emphasise auditability, licensing clarity and deployment tooling.

• Will future flagship models also be open?

Why it matters: Defines the line between free and paid. Expected answer: Undecided; leadership wants a recommendation. Impact on approach: We'll propose an explicit "open core" boundary.

Strategic assumptions:

1. Cohere monetises the platform (North, Compass, Model Vault, support, fine-tuning, indemnity), not the weights.
2. Openness is a trust and distribution lever, particularly for sovereign AI.
3. Cohere keeps some capabilities proprietary to protect differentiation.