Reading up on Together AI

The post shows how to fine‑tune a Jev‑like classifier on Qwen3.5 4B using open datasets for about $17 and deploy it on Together AI.

The cost of handing off a job from a cheap model to a frontier model depends on prompt caching rates, which can make routing more expensive than using the frontier alone.

Open-source models remain costlier to serve reliably due to infrastructure, utilization, and operational overhead, making hosted APIs from OpenAI and Anthropic more economical despite free model weights.

AI inference infrastructure is bifurcating into reserved compute platforms, which monetize predictability, and inference APIs, which monetize utilization through aggregation.

Contra Tim Dettmers, Dan Fu argues current AI systems significantly underutilize hardware, models lag hardware buildout, and multiple paths exist to useful AGI.