Mistral AI drops new open-source model that outperforms GPT-4o Mini with fraction of parameters
1 min read
Summary
French AI start-up Mistral AI has developed a new open-source model, Mistral Small 3.1, which it says outperforms Google and OpenAI models whilst being smaller in size and more efficient.
The firm claims the model processes information at a speed of 150 tokens per second, making it suitable for applications that require rapid response times.
Released under the Apache 2.0 licence, the tool highlights a growing divide in the AI industry between closed, proprietary systems and accessible, open-source alternatives.
Founded in 2023 by former researchers from Google DeepMind and Meta, Mistral AI has quickly established itself as Europe’s leading AI start-up, with a valuation of $6bn.
The company’s focus on algorithmic improvements and training optimisation to maximise capability from smaller architectures addresses the challenge of enormous computational and energy costs associated with state-of-the-art AI systems.
This lightweight approach could become the industry standard as climate concerns and energy costs increasingly impact AI deployment.