How to choose the right open model for production

Together AI · 6 months ago

The article provides guidance on selecting open-source language models for production use, comparing them against closed proprietary models like GPT-5 and Claude. It recommends parameter sizes of 300B+ for high-tier tasks, 70-250B for medium-tier, and under 32B for low-tier work, with specific model families listed across different origins. Success requires defining tradeoffs between cost, speed, and quality, manually evaluating models on representative data, and potentially fine-tuning smaller models when they approach acceptable performance.