holy cow 2109 days !!

There is a new "Model Router" release almost every day these days and we released ours in April '26. Model routers are becoming an important part of the AI stack as they match each request to the model best suited for it. But routing still produces a singular answer from a singular model. So we are releasing another novel way to improve the $/Intelligent Task. @digitalocean's Model Synthesis gets more intelligence from AI, not by choosing a single model but by combining the strengths of several. Multiple models work on the same problem in parallel. A synthesizer then evaluates their responses, reconciles different perspectives, and produces one stronger answer. Model Synthesis is now available in the DigitalOcean Inference Engine, through a single inference call and the early results are compelling. On the DRACO deep-research benchmark, an open-model combination of GLM 5.2 and Kimi K2.6 scored higher than every individual model we tested, at just $0.83 per task. The synthesizer itself proved especially important: choosing the right synthesizer had a greater impact on quality than simply adding more models.Model routing finds the best model for a task.Model synthesis goes further—it assembles the best intelligence for the task. Learn more here: digitalocean.com/blog/model-…