Mixtral 8x22B sets new benchmark for open models

Ryan Daws is a senior editor at TechForge Media with over a decade of experience in crafting compelling narratives and making complex topics accessible. His articles and interviews with industry leaders have earned him recognition as a key influencer by organisations like Onalytica. Under his leadership, publications have been praised by analyst firms such as Forrester for their excellence and performance. Connect with him on X (@gadget_ry) or Mastodon (@gadgetry@techhub.social)

Mistral AI has released Mixtral 8x22B, which sets a new benchmark for open source models in performance and efficiency. The model boasts robust multilingual capabilities and superior mathematical and coding prowess.

Mixtral 8x22B operates as a Sparse Mixture-of-Experts (SMoE) model, utilising just 39 billion of its 141 billion parameters when active.

Beyond its efficiency, the Mixtral 8x22B boasts fluency in multiple major languages including English, French, Italian, German, and Spanish. Its adeptness extends into technical domains with strong mathematical and coding capabilities. Notably, the model supports native function calling paired with a ‘constrained output mode,’ facilitating large-scale application development and tech upgrades.

With a substantial 64K tokens context window, Mixtral 8x22B ensures precise information recall from voluminous documents, further appealing to enterprise-level utilisation where handling extensive data sets is routine.

In line with fostering a collaborative and innovative AI research environment, Mistral AI has released Mixtral 8x22B under the Apache 2.0 license. This highly permissive open-source license ensures no-restriction usage and enables widespread adoption.

Statistically, Mixtral 8x22B outclasses many existing models. In head-to-head comparisons on standard industry benchmarks – ranging from common sense, reasoning, to subject-specific knowledge – Mistral’s new innovation excels. Figures released by Mistral AI illustrate that Mixtral 8x22B significantly outperforms LLaMA 2 70B model in varied linguistic contexts across critical reasoning and knowledge benchmarks:

Furthermore, in the arenas of coding and maths, Mixtral continues its dominance among open models. Updated results show an impressive performance improvement in mathematical benchmarks, following the release of an instructed version of the model:

Prospective users and developers are urged to explore Mixtral 8x22B on La Plateforme, Mistral AI’s interactive platform. Here, they can engage directly with the model.

In an era where AI’s role is ever-expanding, Mixtral 8x22B’s blend of high performance, efficiency, and open accessibility marks a significant milestone in the democratisation of advanced AI tools.

(Photo by Joshua Golde)

See also: SAS aims to make AI accessible regardless of skill set with packaged AI models

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

Tags: , , , , , , ,

View Comments
Leave a comment

Leave a Reply