Every time that Sonnet is acting like it has brain damage (which is once or twice a day), I switch to Opus and it seems to sort things out pretty fast. This is unscientific anicdata though, and it could just be that switching models (any model) would have worked.
This seems like a case of reversion to the mean. When one model is performing below average, changing anything (like switching to another model) is likely to improve it by random chance...
This is a great use case for sub-agents IMO. By default, sub-agents use sonnet. You can have opus orchestrate the various agents and get (close to) the best of both worlds.
In this case I don't think the controller needs to be the smartest model. I use sonnet as the main driver and pass the heavy thinking (via zen mcp) onto Gemini pro for example, but I could use openai or opus or all of them via OpenRouter.
Subagents seem pretty similar to using zen mcp w/ OpenRouter but maybe better or at least more turnkey? I'll be checking them out.
Amp (ampcode.com) uses Sonnet as its main model and has GPT o3 as a special purpose tool / subagent. It can call into that when it needs particularly advanced reasoning.
Interestingly I found that prompting it to ask the o3 submodel (which they call The Oracle) to check Sonnet's working on a debugging solution was helpful. Extra interesting to me was the fact that Sonnet appeared to do a better job once I'd prompted that (like chain of thought prompting, perhaps asking it to put forward an explanation to be checked actually triggered more effective thinking).
Is there a way to get persistent sub-agents? I'd love to have a bunch of YAML files in my repository, one for each sub-agent, and have those automatically used across all Claude Code instances I have on multiple machines (I dev on laptop and desktop), or across the team.
In my experience the best use for subagents is saving context.
Example: you need to review some code to see if it has proper test coverage.
If you use the "main" context, it'll waste tokens on reading the codebase and running tests to see coverage results.
But if you launch an agent (a subprocess pretty much), it can use a "disposable" context to do that and only return with the relevant data - which bits of the code need more tests.
Now you can either use the main context to implement the tests or if you're feeling really fancy launch another sub-agent to do it.
AFAIK subagents inherit the default model since v1.0.64. At least that's the case for me with the Claude Code SDK — not providing a specific model makes subagents use claude-opus-4-1-20250805.
I have suspected for a long time that hosted models load shed by diverting some requests to lesser models or running more quantized versions under high load.
