Mistral AI has announced the release of two new models designed for on-device computing and edge use cases: Ministral 3B and Ministral 8B, collectively known as "les Ministraux."
On the first anniversary of Mistral 7B model, Mistral AI has set a new benchmark in the sub-10B category for knowledge, commonsense reasoning, and efficiency. These models support up to 128k context length and are poised to transform various sectors, from autonomous robotics to local analytics.
The models are designed to provide compute-efficient and low-latency solutions for critical applications requiring privacy-first inference.
Mistral AI's internal benchmarks show les Ministraux consistently outperforming peers across multiple tasks. Notably, the Ministral 3B model surpasses the performance of the larger Mistral 7B on most benchmarks, highlighting significant progress in AI model efficiency.
Both models are available immediately, with Ministral 8B priced at $0.1 per million tokens and Ministral 3B at $0.04 per million tokens for API access. The company is also offering commercial licenses for self-deployed use and assistance with lossless quantisation for specific use-cases.