Mistral: Mistral Small 3.2 24B

Mistral-Small-3.2-24B-Instruct-2506 is a powerful and updated 24B parameter model from Mistral, specifically optimized for superior instruction following, significant repetition reduction, and enhanced function calling capabilities. This version (3.2) shows substantial improvements over its 3.1 predecessor, delivering higher accuracy on WildBench and Arena Hard, effectively reducing infinite generations, and making significant gains in tool use and structured output tasks. This versatile model supports both image and text inputs, providing structured outputs for complex tasks. It excels in function/tool calling and demonstrates strong performance across various benchmarks, including coding (HumanEval+, MBPP), STEM (MMLU, MATH, GPQA), and vision tasks (ChartQA, DocVQA). With a generous 131K token context window and a 4K token max output, it's ideal for chat, code, and translation. Access this FREE model on Multi AI, with competitive pricing at $0.06/$0.18 per 1M input/output tokens.