So, I recently started using the Hermes agent and hooked it up with Step 3.5 Flash. It’s a cheap and fast model, so I thought it would be enough for my workflow. However, I soon found that it hallucinates during a lot of tasks, and the output quality isn’t something to write home about.

On the other hand, using an expensive or frontier-level model means I’ll burn through cash fast. I went down the rabbit hole to find a solution and discovered I could solve this using a model router. Dedicating a capable model like Kimi K3 for complex tasks while Step handles the busywork.

Initially, I tried LiteLLM, but I noticed a significant delay in the routing process. I thought that latency was just a sacrifice I had to make. Then, while browsing GitHub one day, I came across Bifrost and decided to give it a shot. Honestly, it is super fast, and I haven’t faced any of the delay issues that were prevalent in LiteLLM.

So my question is: is using a model router the most practical choice when running Hermes or OpenClaw? What do you guys usually do?