Cohere’s smaller, faster R-Series model excels at the RAG, speaking in 23 languages

Proven to support a wide range of enterprise use cases – including those that are not cost-prohibitive or resource-intensive Large linguistic models (LLMs) — Artificial Intelligence startups cohere Command has released the R7B, the smallest and fastest in the R model series.

The Command R7B is designed to support rapid prototyping and iteration and uses Retrieval Augmented Retrieval (RAG) technology to improve its accuracy. The template has a context length of 128KB and supports 23 languages. Cowher says it outperforms others in its class of open-weight models — Google’s Gemma, Meta’s Llama, and Mistral’s Ministral — on tasks including mathematics and programming.

“The model is designed for developers and businesses that need to improve speed, cost performance, and compute resources for their use cases,” Cohere co-founder and CEO Aidan Gomez said. He writes in a blog post Announcing the new model.

Outperform competitors in mathematics, programming and RAG

Cohere has been strategically focused on enterprises and their unique use cases. Company provided Driving-R in March And the strong R+ command In April, upgrades were made All year round To support speed and efficiency. It has teased the Command R7B as the “final” model in its R series, and says it will release model weights to the AI ​​research community.

Cowher noted that a critical focus area when developing the R7B command was improving performance in mathematics, reasoning, code and translation. It seems that the company has succeeded in these areas, with the new smaller model topping the list HuggingFace Open the LLM Leaderboard Against similarly sized open weight models including the Gemma 2 9B, Ministral 8B and Llama 3.1 8B.

Furthermore, the smallest model in the R Series outperforms competing models in areas including AI factors, tool utilization and RAG, helping improve accuracy by anchoring model output to external data. Cowher says the Command R7B excels at conversational tasks including technical workplace and enterprise risk management (ERM) assistance; Technical facts; Media workplace and customer service support; Human Resources FAQs; And summarizing. Cowher also notes that the model is “exceptionally good” at retrieving and processing digital information in financial settings.

Finally, the Command R7B ranked first, on average, in important criteria including Instruction Follow-Up Evaluation (IFeval); Big Hard Bench (BBH); Google-proof Graduate Level Questions and Answers (GPQA); Soft, multi-step thinking (Moore); and Understanding language is a huge multitasker (Mello).

Remove unnecessary calling functions

The R7B command can use tools including search engines, APIs, and vector databases to extend its functionality. Cohere reports that using the model tool performs strongly against competitors on the Berkeley Function Calling Leaderboard, which evaluates a model’s accuracy in calling functions (calling to data and external systems).

Gomez points out that this proves effective in “diverse and dynamic real-world environments” and eliminates the need for unnecessary communication functions. This could make it a good choice for building “fast and capable” AI agents. For example, Cowher points out that when acting as an Internet-enhanced search agent, Command R7B can break down complex questions into subobjectives, while also performing well with advanced reasoning and information retrieval.

Due to its small size, Command R7B can be deployed on low-end and consumer CPUs, GPUs, MacBooks, allowing on-device inference. The model is now available on Cohere and HuggingFace. The price is $0.0375 per million input tokens and $0.15 per million output tokens.

“It is an ideal choice for organizations looking for a cost-effective model that centers their internal documentation and data,” Gomez wrote.

Daily insights into business use cases with VB Daily

If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from organizational transformations to hands-on deployments, so you can share insights to maximize ROI.

Read our privacy policy

Thanks for subscribing. Check more VB Newsletters are here.

An error occurred.