AI News·4 min read

Cohere Releases Command A+ — First Apache 2.0 Licensed Frontier AI Model

Cohere launches Command A+, a 218-billion-parameter open-source model with lossless 4-bit quantization and native citations under the permissive Apache 2.0 license.


Cohere Just Changed the Open-Source AI Game — Here's How

Cohere, the Canadian AI lab co-founded by "Attention Is All You Need" co-author Aidan Gomez, has released Command A+ under a full Apache 2.0 license. This is the first time a frontier-grade AI model with this level of capability has been made fully open source with such a permissive license.

Why Is Apache 2.0 Licensing a Big Deal?

Most "open" AI models come with restrictive licenses that limit commercial use or require attribution. Apache 2.0 allows anyone to use, modify, and distribute the model commercially with minimal restrictions. This means enterprises can deploy Command A+ in production without legal uncertainty.

What Makes Command A+ Technically Special?

The model uses a Sparse Mixture-of-Experts architecture with 218 billion total parameters but only 25 billion active during any generation step. The real breakthrough is its 4-bit quantization (W4A4), which compresses the model with near-zero quality loss. It can run on a single NVIDIA B200 GPU or just two H100s — making frontier AI accessible without massive infrastructure.

How Fast Is It?

Command A+ achieves 375 tokens per second with a time-to-first-token of just 113 milliseconds. That represents up to 63% faster output and 17% lower latency compared to Cohere's previous Command A Reasoning model, despite being significantly larger.

What Are Native Citations and Why Do They Matter?

Command A+ includes built-in citation capability, meaning it can reference source documents in its responses. This is critical for enterprise use cases like legal research, medical analysis, and financial reporting where verifiable sources are non-negotiable.

Frequently Asked Questions

Q: Can I run Command A+ on consumer hardware? A: The 4-bit quantized version requires at minimum two H100 GPUs or one B200. It is designed for enterprise deployment rather than consumer devices.

Q: How does it compare to GPT-5.5 or Claude Opus 4.7? A: While proprietary models may edge ahead on some benchmarks, Command A+ offers competitive performance with full transparency and no vendor lock-in.

Q: What is "sovereign AI"? A: Sovereign AI refers to the ability to run and control AI models entirely within your own infrastructure, without depending on external API providers. Command A+ enables this.


Stay ahead of the AI curve. Follow @AiForSuccess for daily insights.

📬 Want more AI solopreneur insights?

Subscribe to our weekly newsletter →
☕ Enjoy this article? Support the author

Related Articles