Hey everyone,
I wanted to share a workflow I’ve been experimenting with that might be interesting for the community.
Right now, one of the most advanced models available is (Claude SONNET 4.5. It’s widely recognized as being on par with something like GPT-5 High in terms of performance, but of course, it comes with a pretty steep cost. On the other end of the spectrum, there’s GROK CODE FAST. Despite my personal reservations about Elon Musk, I have to admit it’s impressively fast and fairly coherent, though it does make a fair amount of mistakes.
So, how do I balance performance with limited resources?
I use them in a tag team strategy:
-
SONNET acts like the wise senior developer: it does the planning, strategy, and creates structured guidelines.
-
GROK plays the role of the fast junior dev: it executes the plan, sometimes making mistakes, but usually handling the heavy lifting quickly.
The workflow is file-based: I let SONNET produce “blueprints” in document form (usually markdown), which I then feed to GROK to carry out. Surprisingly, this works quite well.
For example, I had a large Streamlit app stuck in a single massive file. I asked SONNET to design a refactoring plan to split it into modules. SONNET generated a detailed guideline document, then I passed that to GROK, which processed and completed the refactor successfully.
So far, this hybrid approach is saving me money (i have the base subscription) while still letting me take advantage of SONNET’s strategic depth and GROK’s speed.
Curious if anyone else here has tried mixing models in a similar way — and if so, what workflows you’ve found effective?
