- We do something similar to OpenRouter which measures the latency of the different providers, to ensure we always get the fastest results

- Users can cancel a single model stream if it's taking too long

- The orchestrator is pretty good at choosing what models for what task. The actual confidence scoring and synthesis at the end is the difficult part that you cannot do naively, however, the orchestrator plays the biggest part in optimizing cost + speed. I've made sure that we don't exceed 25% extra in cost or time in the vast majority of queries, compared to equivalent prompts in ChatGPT/Gemini/etc.

The reason why this is viable IMO is because of the fact that you can run multiple less-intelligent models with lower thinking efforts and beat a single more-intelligent model with a large thinking effort. The thinking effort reduction speeds up the prompt dramatically.

The sequential steps are then:

1. Ensemble RAG 2. Orchestrator 3. Models in parallel 4. Synthesizer

And retries for low-confidence (although that's pretty optimized with selective retries of portions of the answer).

I.e. you get 3 replies. 80% confidence. You decide at 80% you are fairly good but happy to wait 5 seconds for completion / 500ms for time to first token. If either breaches you give the current answer.

But if you are at 5% you wait for 60s total/2s for a token since the upside of that unspoken model is much higher.

Basically wagering time for quality in a dynamic prediction market in front of the LLM.

Maybe this insight is why I feel hallucinations are much rarer in the last 12 months on top models. Are they being detected before they get sent out.

Would love to hear how Sup works out for you.