French AI startup Mistral AI has unveiled Mistral Small 3.1, a new open-source model that processes text and images using just 24 billion parameters while matching or exceeding the performance of larger proprietary models. Running at 150 tokens per second with a 128k token context window, the model demonstrates remarkable efficiency compared to industry giants. Released under Apache 2.0 license, it represents a strategic departure from competitors' closed systems. The $6 billion startup's approach emphasizes European digital sovereignty and sustainable AI development, backed by partnerships with Microsoft, AFP, and various European institutions.
French AI startup Mistral AI has unveiled Mistral Small 3.1, a new open-source model that processes text and images using just 24 billion parameters while matching or exceeding the performance of larger proprietary models. Running at 150 tokens per second with a 128k token context window, the model demonstrates remarkable efficiency compared to industry giants. Released under Apache 2.0 license, it represents a strategic departure from competitors' closed systems. The $6 billion startup's approach emphasizes European digital sovereignty and sustainable AI development, backed by partnerships with Microsoft, AFP, and various European institutions.