Cohere, the Toronto, Canada-based startup founded by former Google AI researchers and co-authors of Transformer paper, announced their groundbreaking new AI large language model (LLM), Command R+. This LLM is tailored specifically for enterprise adoption.
Command R+ also published metrics showing its superiority over competing AI products such as OpenAI’s GPT-4 (which powers ChatGPT). On various benchmark tests for AI systems.
Since then, Command R+ has been made available through Microsoft Azure cloud service and most recently Amazon Sagemaker — their fully managed machine learning development platform.
On Azure, accessing Command R+ costs $0.015/$0.003 per 1000 tokens out/in (the cost for getting the model to provide its outputs or entering prompts into it respectively – tokens roughly correspond to words or parts of words and are represented to LLM as numbers that represent their physical position within space). Open source developer Simon Willison has an amazing interactive example here).
Differing Price Structures
Sagemaker offers various pricing structures ranging from $82.89 to $1469 per hour when used for realtime inference, meaning running it on end users (not for development) in production environments.
Microsoft made an instant announcement via social networking site X that Command R+ was “the first available on Microsoft Azure,” as well as writing an in-depth blog post stating its cloud service now offers over 1,600 foundation models from Databricks, Deci AI, Hugging Face Meta Microsoft Research Mistral AI NVIDIA OpenAI Stability AI
“Command R+ is designed for enterprises looking to use internal data and documents from internal systems for creating tailored, accurate language models,” according to John Montgomery, Corporate Vice President for Program Management of AI Platform at Microsoft. When coupled with Cohere Embed and Rerank (which will soon become part of Microsoft’s model catalog), Command R+ delivers outstanding accuracy for AI applications requiring information from documents or enterprise data sources.
At present, Azure appears to be competing fiercely with other cloud services for AI model offerings, and seems determined to make itself the “go-to” destination for enterprise developers looking for AI models and hosting spaces. After investing and being the first to offer Mistral Large from French AI startup Mistral; as well as supporting OpenAI over time; Microsoft appears intent on becoming the “one stop shop” for enterprise developers looking for AI solutions and host locations.
Amazon, too, has been steadily building out their model library on Bedrock — their LLM-focused development platform — adding Anthropic’s Claude 3 model family shortly thereafter. So the presence of Command R+ on AWS Sagemaker stands out, particularly considering it is not available through Bedrock.
As part of their fiercely competitive ongoing battle, the two largest cloud providers by market share are adding more AI models as part of a highly strategic effort to remain ahead of rival providers. It makes sense as adding more models gives cloud providers access to a wider range of use cases and customer needs to address, not to mention serving up more bragging rights than smaller mom and pop shops that only offer limited options.
Cohere’s business may benefit from becoming available on more major cloud and AI development platforms, offering potential customers more ways to access it and familiarize themselves with it. According to The Information, revenue has been slow in coming in while OpenAI has seen good success expanding enterprise user subscriptions by 600,000. Bloomberg states this growth was achieved with enterprise customers being prioritized compared to consumers which is what Cohere was created for and remains focused on winning over.
Why doesn’t Command R+ exist on Bedrock?
Aidan Gomez, Cohere’s CEO and cofounder, noted its absence on X, posting: “Excited that Command R+ has finally made it onto AWS Cloud Sagemaker! Please contact your AWS rep and request they add it along with Command R, Rerank and R+ to Bedrock!!”
Reaching out to AWS contacts regarding why Command R+ wasn’t yet available on Amazon Bedrock and when we could expect its introduction was met with this response: AWS does not publicly reveal its product roadmap; however, as the only cloud provider to offer the most popular and advanced foundation models to customers. Cohere plays an integral role in our work here and we look forward to continuing our relationships.
