Third-party cyber evaluations involving OpenAI models
openai.com·14h ago
Cloud & Infrastructure
The New Stack·Jun 15·5 min read
Cohere sold sovereign AI to enterprises, now it’s targeting developers with its first coding model
TL;DR
Enterprises, especially in regulated sectors like finance and healthcare, needed AI solutions that kept their data secure and under their control. Cohere has developed North Mini Code, its first coding model, which is available under the Apache 2.0 license.
✦ Why It Matters
Developers can now access a secure coding model that supports data sovereignty while fostering innovation.
Key Takeaways
- Explore North Mini Code for self-hosted AI solutions.
- Evaluate performance metrics against existing coding models.
- Consider the implications of open-weight models for your projects.
- Assess infrastructure needs for deploying AI models effectively.
Full Summary
Cohere, a Canadian foundation model company, has shifted its focus from selling AI solutions to enterprises to targeting developers with its new coding model, North Mini Code. This model features 30 billion parameters, with 3 billion active, and is optimized for complex coding tasks.
It can run on a single Nvidia H100 GPU, making it accessible for self-hosting or via API. Cohere claims North Mini Code outperforms similar models, achieving a score of 33.4 on the Artificial Analysis Coding Index.
The launch aligns with a trend among companies like Mistral and JetBrains, which are also offering open-weight coding models, allowing developers more control over their AI infrastructure. Cohere's move to an Apache 2.0 license for North Mini Code reflects a broader push for sovereignty in AI development, enabling organizations to manage their own data and infrastructure.