Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now
Earlier this week, Cohere, the Toronto, Canada-based startup co-founded by former Google AI researchers and a Transformer paper co-author, unveiled its powerful new AI large language model (LLM) geared for enterprise adoption: Command R+.
It also posted metrics showing how Command R+ surpassed competitor OpenAI’s GPT-4 (which powers ChatGPT) on common AI benchmarking tests.
Since then, Command R+ has been made available on Microsoft’s Azure cloud service and just today, Amazon Sagemaker, the rival company’s fully managed machine learning development platform.
On Azure, accessing Command R+ costs $0.015/$0.003 per 1000 tokens out/in (the price for getting the model to respond with its own outputs and for entering in prompts, respectively. Tokens are roughly applicable to words or parts of words, represented to the LLM as numbers that imply where that word or concept sits in a 3D representation of space relative to other related concepts. Open source developer Simon Willison has a great interactive example here).
AI Scaling Hits Its Limits
Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:
- Turning energy into a strategic advantage
- Architecting efficient inference for real throughput gains
- Unlocking competitive ROI with sustainable AI systems
Secure your spot to stay ahead: https://bit.ly/4mwGngO
Differing pricing schemes
On Sagemaker, it is billed differently at $82.89 to $146.93 per hour for running it in realtime inference (using the model in production for end users, not for development).
Microsoft was quick to boast on the social network X about how Command R+ was “available first on Microsoft Azure” and put out a lengthy blog post noting how its cloud service “now offers more than 1,600 foundation models (LLMs and SLMs) from Databricks, Deci AI, Hugging Face, Meta, Microsoft Research, Mistral AI, NVIDIA, OpenAI, Stability AI.”
“Command R+ is built for enterprises that plan to leverage their internal data and documents for building tailored and accurate language models,” wrote John Montgomery, Microsoft’s Corporate Vice President, Program Management, AI Platform at Microsoft. “When paired with Cohere Embed and Rerank which will be available in the model catalog soon, Command R+ offers leading accuracy for advanced AI applications requiring information from documents and enterprise data sources.”
Fiercely competitive market in cloud for AI model offerings
Coming after its move to invest in and be the first to offer the proprietary Mistral Large from the heretofore open source French AI startup of the same name, and amid its longstanding support and investment of OpenAI, Microsoft seems to be intent on making Azure the “one-stop shop” for any enterprise developer looking to choose between the variety AI models and places to host them.
Yet Amazon, too, has been steadily building out its model library on Bedrock, its LLM-focused development platform, and was quick to add Anthropic’s Claude 3 model family to it earlier this year. So the fact that Command R+ is available on AWS Sagemaker is interesting and notable, especially when it’s not on Bedrock.
The move to add more models generally by the two largest cloud providers by market share is part of their fiercely competitive ongoing battle. It makes sense in that the more AI models the cloud provider can offer, the more potential use cases and customer needs it can attempt to address, as well as just be a general bragging right. It’s like being a supermarket with numerous brands as opposed to a small mom and pop shop with just a few choices.
For Cohere, getting placement on more leading cloud and AI development platforms is presumably better for business, too, offering potential customers more ways to access it and become familiar with the name. Cohere is also reportedly slow to generate revenue, according to The Information, while OpenAI has apparently had decent success growing its enterprise user subscribers to 600,000, according to Bloomberg — the very customers Cohere was founded to reach and still focused on winning over.
Why isn’t Command R+ on Bedrock?
Even Aidan Gomez, Cohere’s CEO and cofounder, noted the omission on X, posting: “Excited to have Command R+ up on @awscloud Sagemaker. Please email your local AWS rep and ask for it and Command R and Rerank to be added to Bedrock!”
We reached out to AWS contacts to ask about why Command R+ wasn’t yet available on Amazon Bedrock and when it could be expected, and a spokesperson responded with the following statement: “AWS does not disclose product roadmap; however, AWS is the only cloud provider to offer the most popular and advanced foundation models to customers. Model providers, like Cohere, are integral to this work, and we look forward to continuing our relationships with them.”
