Mistral Small 3

Mistral Small 3 is a 24B-parameter language model released under the Apache 2.0 license. It offers performance comparable to larger models like Llama 3.3 70B while being more than 3x faster on the same hardware. Designed for local deployment, it excels in tasks requiring robust language and instruction following with very low latency. The model can be quantized to run on a single RTX 4090 or a Macbook with 32GB RAM

• 24B parameters
• Apache 2.0 license
• Low latency (150 tokens/s)
• 81% accuracy on MMLU
• 32k context window
• Multilingual support
• Function calling capabilities
• Optimized for quantization

• Fast-response conversational assistance
• Low-latency function calling
• Fine-tuning for subject matter experts
• Local inference for sensitive data
• Fraud detection in financial services
• Customer triaging in healthcare
• On-device command and control in robotics and manufacturing
• Virtual customer service
• Sentiment and feedback analysis